awesome-ai-tools
agent
Fail
Health Warn
- No license — Repository has no license file
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 27 GitHub stars
Code Fail
- rm -rf — Recursive force deletion command in .astro/data-store.json
- network request — Outbound network request in .astro/data-store.json
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
A curated, categorized reference of AI-powered coding tools. Covers full IDEs, editor extensions, terminal agents, autonomous agents, browser-based builders, and code review platforms.
README.md
AI IDEs & Coding Assistants
🌐 Browse the interactive directory: ai.dosa.dev: search, filter, compare, and read per-tool reviews (pricing, features, verdicts).
📣 Advertise your tool: the directory ships a site-wide Featured slot for sponsors: reach developers actively evaluating AI coding tools. Rates and details at ai.dosa.dev/advertise.
Manually curated, enhanced w/ Devin
A curated, categorized reference of AI-powered coding tools as of September 2026.
Covers full IDEs, editor extensions, terminal agents, autonomous agents, browser-based builders, and code review platforms.
225+ tools across 10+ categories.
Important: Do not open a pull request to submit a new tool. Use the Tool Submission form instead, it runs the automated pipeline that updates all data files.
📋 Table of Contents
- 🖥️ AI-Native IDEs & Editors
- 🔌 IDE Extensions & Plugins
- 💻 Terminal & CLI Agents
- 🤖 Autonomous & Async Agents
- 🌐 Browser-Based & App Builders
- 🛡️ AI Code Review & Security
- 🧰 General-Purpose AI Assistants
- 🛠️ Developer Productivity & Workflow
- 🖊️ Editor Platforms with Native AI Features
- 🔧 Hardware & Edge AI
🖥️ AI-Native IDEs & Editors
Full standalone editors built from the ground up with AI at the core.
| Tool | Company | Notes |
|---|---|---|
| Cursor | Anysphere | VS Code fork; agent mode for multi-file edits; most popular AI-native IDE |
| Devin | Cognition | Autonomous AI software engineer by Cognition; Devin Desktop (formerly Windsurf) is its AI-first IDE. |
| Trae | ByteDance | Free AI IDE (VS Code-based); Builder Mode; GPT-4o + Claude access at no cost |
| Zed | Zed Industries | High-performance multiplayer editor with built-in AI; by creators of Atom |
| PearAI | PearAI | Open-source AI code editor; VS Code fork |
| Antigravity | Dual-mode IDE: Editor View + Manager Surface for orchestrating autonomous agents | |
| Void | Void | Work on Void is currently paused |
| Verdent | Verdent | An agentic coding tool focused on parallel agents and isolated workspaces |
| Pochi | Pochi | Pochi is a VS Code–native AI coding agent built by TabbyML that stands out primarily through its parallel agent execution, local model support, and more. |
| Qoder | Qoder | Agentic Coding Platform for Real Software. |
| Z Code | Z | Z Code combines the best AI agents with your existing tools so you can plan, code, review, and deploy without friction. |
| toad | Batrachian AI | A unified interface for AI in your terminal.; GitHub |
| Nimbalyst | Nimbalyst | Open-source visual workspace for building with Codex, Claude Code, and more. Manage agents, sessions, tasks, files. Visually edit markdown, mockups, diagrams, diffs, and code.; GitHub |
| Superset | Superset | Code Editor for the AI Agents Era - Run an army of Claude Code, Codex, etc. on your machine |
| archestra | Archestra Inc. | Enterprise AI Platform with guardrails, MCP registry, gateway & orchestrator |
| codex-profiles | Chai Pin Zheng | Switch Codex CLI and Desktop accounts with isolated CODEX_HOME profiles |
| Reasonix | ESEngine | A DeepSeek-native coding agent, for your terminal. |
| MiniMax Code | MiniMax | Coding agent powered by MiniMax M-series models: an open-source (MIT) terminal CLI (mcode, installed via one-command script or npm install -g @minimax-ai/code) that builds, tests and iterates from a prompt with MiniMax or your own Anthropic/OpenAI-compatible model, plus desktop, web and mobile apps that remember your habits, build Agent teams and automate repetitive work; GitHub |
| Meta Muse | Meta | An agent for your most complex coding workstreams. Build, debug and ship with Muse Code. |
| DeepCode | HKUDS | DeepCode: Open Agentic Coding (Paper2Code & Text2Web & Text2Backend) |
| Fletch | FWDAI | Open source multi-agent runner for your coding agents that runs then in parallel and in real isolation (Docker or Seatbelt). Allows chaining them into deterministic workflows that plan, build, review and test. You ship AI-written code you can actually trust, not just more of it.; GitHub |
| Factory AI | Factory | Complete software development agents for individuals. |
| ante | AntigmaLabs | Ghost in your shell. Ante is a self-contained agent harness with a highly optimized core. It works like Claude Code or Codex, with none of their dependencies or model constraints.; GitHub |
| unsloth | unsloth | Unsloth lets you run, train, and deploy AI models locally, with support for all types of models.; GitHub |
| bb | bb | bb is an agentic IDE that can control itself. You can seamlessly orchestrate all of your favorite coding agents together and have them programmatically use bb too.; GitHub |
| OpenChamber | OpenChamber | OpenChamber is an open-source workspace for running, supervising, and reviewing AI coding work across desktop, browser, editor, and mobile.; GitHub |
| Waku | EGOIST | Waku is a fast, native desktop app for working with local coding agents. It is built in Rust with GPUI and keeps projects, sessions, transcripts on your machine.; GitHub |
| jcode | Jeremy Huang | jcode is built to be as performant and resource efficient as possible. Every metric is optimized to the bone, which is important for scaling multi-session workflows. Here we sample a few metrics to show the difference: RAM usage and boot up.; GitHub |
| Grok Bot | X | AI teammates you can give real work to. Bots can sign in to your tools, use them just like you do, and come back with finished work. |
| Ori | OpenRouter | Your favorite harness with every model. |
| AgentOne | BestCodes | A free AI agent and deep-researcher.; GitHub |
| OneCLI | OneCLI | Open-source sandboxed agent harness for teams. Giving every employee a secured personal agent.; GitHub |
| Termy | Lasse | The terminal, at full speed ⚡; GitHub |
| Paseo | getpaseo | Orchestrate multiple coding agents from desktop and mobile.; GitHub |
| OpenWorker | Andrew Ng | OpenWorker is an open-source AI coworker that lives on your desktop and delivers finished work, not just chat: a polished document, a Slack reply with the numbers, an updated calendar, a triaged inbox. |
| MonoCode | Nick | A desktop UI for your coding agents.; GitHub |
| Emdash | generalaction | Emdash is an open-source agentic development environment for running multiple coding agents in parallel, each isolated in its own Git worktree. It supports local and remote SSH projects and works with Claude Code, Codex, Cursor, OpenCode, Amp, Devin, and other provider CLIs.; GitHub |
| Atlarix | NorahLabs | Private AI workstation: a native desktop app (macOS/Linux/Windows) with Explore/Plan/Build/Debug/Review modes. Bring any model via BYOK or run fully local (Ollama/LM Studio), or use its managed Core tier; code never leaves the machine, an OS-level sandbox and hunk-level approval gate every edit, and bundled-ripgrep search needs no index. |
| Atomic Chat | Atomic Local Ai | Atomic is an open-source AI agent optimized for local models. |
| TasteCode | Leon Lin | A local desktop workspace for AI coding agents, with Design Mode built in. |
| Orca | Stably AI | Open-source (MIT) "agent development environment" (ADE): runs Claude Code, Codex, OpenCode, Cursor and 25+ other CLI agents in parallel, each in its own Git worktree, with terminals, diffs, a built-in browser and an orca CLI in one desktop app (macOS/Windows/Linux); fan one prompt across agents and merge the winner; iOS/Android companion for remote steering.; GitHub |
| Orchestrator.inc | Orchestrator.inc | Open-source (Apache-2.0) desktop workspace (macOS, Windows, Linux, plus iPhone and Android apps) for planning, running and supervising teams of coding agents from task to merged PR: each worker gets one task, one agent and its own branch and worktree; a persistent project orchestrator plans work, splits it into tasks and delegates to workers; a live Kanban (Working, Needs you, In review, Ready to merge) is derived from session, PR, CI and review state; per-worker terminal, structured Chat, agent-controllable isolated browser and PR/CI/review feedback routed back to the owning agent; supports Claude Code, Codex, Cursor, opencode, Aider, GitHub Copilot, Amp, Cline, Goose, Devin, Kiro and 30+ agents; GitHub |
| Supercode | Yash Dewasthale | The open source harness agent. |
| Rune | Unstable Build | Open-source (GPL-3.0) GPU-accelerated native IDE and terminal multiplexer for macOS/Linux (OpenGL/Metal, no Electron): nine workspace slots with unlimited terminals/tabs/windows, built-in language intelligence and debugging, a built-in coding agent plus IDE-grade skills for external agents, e2e-encrypted peer network (headscale) to reach your instances from any machine, and pkg install extensions from Git repos; GitHub |
| Kimi Code | Kimi | Kimi Code is now available on desktop, CLI, and IDE plugins engineered to drop into any dev workflow, answer faster and more reliably, and get programming tasks done in record time. |
| Step Code | stepfun-ai | Step Code runs in your terminal and handles the full task loop🤖reading code, making changes, and running tests. |
| Athas | Athas | Lightweight open-source (AGPL-3.0) cross-platform code editor built with Tauri (Rust + React) that hosts coding agents inside the editor: Claude Code, Codex, Gemini CLI, GitHub Copilot CLI, OpenCode, Qwen Code and Kimi CLI run as ACP/app-server sessions with permission prompts, message queues and chat history, alongside chat with Anthropic, OpenAI, Google, DeepSeek, Mistral, Grok, OpenRouter or Ollama models. Also ships Git, LSP, vim keybindings, integrated terminal, database viewers, collaboration and enterprise managed mode; brew install --cask athas / winget install athasdev.Athas.; GitHub |
| ccgui | zhukunpenglinyutong | Open-source (MIT) multi-engine AI coding desktop client (Tauri 2 + React + Rust, macOS/Windows/Linux) that puts Claude Code, Codex CLI, Kimi CLI, Grok CLI, Pi, OMP, DeepSeek Harness, Antigravity, OpenCode and Qoder behind one GUI via native protocol adapters: streaming output with thinking traces and tool-call rows, session history/resume, per-tab model and effort overrides, provider-channel presets (imports from CC Switch), plus a file tree with CodeMirror editor, PTY terminal, Git panel, plugin SDK and LAN web access; bilingual Chinese/English; GitHub |
| Zeron | Zeron | Open-source (MIT) native control plane for coding agents: runs Claude Code, Codex, Cursor, Devin, Grok, Hermes, Pi and Antigravity sessions on your own machines and lets you steer them from one fast native window on desktop, phone or another device; local-only by default (no account), optional multi-device sync via a small per-device engine daemon so an always-on VPS keeps agents working after you close your laptop; live branch diffs, terminals and worktrees per session; Linux, macOS and Windows; GitHub |
| Juggler | Julian Storer | Open-source (AGPL-3.0; extension SDK Apache-2.0) visual workbench for AI coding agents by the creator of JUCE: sessions are persistent, branchable Yjs trees shown in Miller columns, with inspectable tool calls and raw LLM transactions, editable context and nested sub-threads. Go backend with native desktop app plus headless server so several clients (desktop, browser, phone) attach to one live session; vendor-neutral BYOK for Claude Code, Codex, Copilot, Gemini, Ollama and others; GitHub |
| Whiteboard | /dev/fast (YC W26) | Open-source (MIT) desktop app for reviewing and designing what coding agents build: Claude Code, Codex and other agents get an SDK to draw sequence and entity-relationship diagrams of their changes on a shared canvas, each element linking to the underlying code in a vendored Code OSS editor (VS Code keybindings + LSP); Rust AST-aware semantic diff viewer summarises large functions as pseudocode and folds tests/docs (WASM plugins), plus a Decision Log linking agent traces to requirements; runs locally against your checkouts on macOS, Windows and Linux; GitHub |
| Proliferate | Proliferate | Open-source (AGPL-3.0), self-hostable AI IDE that runs Claude Code, Codex, OpenCode, Cursor and Grok through their native harnesses in parallel, each task in its own Git worktree with its own terminal, conversation and review state; subagent delegation across harnesses, shared MCPs/skills/computer- and browser-use integrations, and reusable scheduled or event-driven workflows with human approval gates; local use is free, optional cloud sandboxes and model gateway are metered, and the full control plane self-hosts on Docker, AWS, GCP, Azure, Kubernetes or air-gapped; GitHub |
| T3 Code | Ping.gg | Open-source (MIT) "agent harness control surface" by Theo and Julius: one web, Electron desktop and iOS/Android app that drives the coding-agent CLIs already on your machine (Codex, Claude Code, Cursor, Grok Build, OpenCode) with your own subscriptions; every thread works on its own branch/worktree, inline diff review and one-button commit/push/PR with generated titles, bodies and changelogs, permission modes and remote access from a phone; npx t3@latest or desktop releases.; GitHub |
| Voidleap Code | Voidleap | Voidleap Code is an agentic IDE designed for token-efficient software development. Mix model providers without locking into one, and track usage through a built-in dashboard. |
| Helicon | Harjot Singh Rana | Free, open-source (MIT) desktop app and VS Code extension for Meta's Muse Code CLI: threads from every project in one sidebar, inline diffs, one-at-a-time approvals and per-thread usage costs. Unofficial, not affiliated with Meta.; GitHub |
| maki | Tony Solomonik | An efficient AI coding agent extendable by neovim-like Lua plugins. |
| Unstoppable AI | Unstoppable AI | Unstoppable AI is an agent IDE that brings Claude Code, Codex and Cursor together with your existing subscriptions. It supports parallel agents and multi-agent workflows, with a terminal and browser organized around each task. |
| ThinkRail | JetBrains | Open-source (Apache-2.0) agentic IDE from the JetBrains InnovationHub, built around the pi coding agent: open a git repo as a project and cut isolated git-worktree workspaces where several concurrent pi chats (per-session model, live token and cost readout, steering, queued follow-ups, subagents, shared editable plans) work beside Monaco editor tabs, a Files tree with git change marks, terminals, diffs and anchored review comments; a graph of living SPEC.md files gives the agent durable project knowledge via spec_* tools, and it reads existing Agent Skills, connects MCP servers with approval prompts, and opens or merges GitHub PRs through your gh; sign in with JetBrains AI or any pi provider (no ThinkRail account); desktop app for macOS Apple Silicon, Windows x64 and Linux, or a thinkrail CLI that serves the same UI in the browser; GitHub |
| Volt | LeulAria | Open-source (MIT) agentic workspace built on the VS Code core. Runs Claude Code, Codex and other ACP agents or local Ollama models, with plan mode, subagents, MCP, Git checkpoints and hunk-level review of agent edits.; GitHub |
🔌 IDE Extensions & Plugins
Plug-in assistants that enhance your existing editor (VS Code, JetBrains, Neovim, etc.).
| Tool | Company | Notes |
|---|---|---|
| GitHub Copilot | GitHub | Industry standard; inline autocomplete + agent mode; 15M+ devs |
| Junie | JetBrains | The JetBrains AI agent; deep integration with IntelliJ, PyCharm, WebStorm |
| Gemini Code Assist | Gemini-powered; free for individuals; VS Code + Google Cloud integration | |
| Amazon Q Developer | AWS | Evolution of CodeWhisperer; multi-file agents; deep AWS integration |
| Tabnine | Tabnine | Privacy-first; on-premise options; all major IDEs; ethically sourced training |
| Cline | Cline | Autonomous VS Code agent; Plan + Act modes; BYOM (zero markup) |
| Kilo Code | Kilo Code | Open-source; 500+ models via OpenRouter; 4 structured modes (Architect/Code/Debug/Orchestrator) |
| Roo Code | Roo Code | VS Code extension; strong on large multi-file changes; fork of Cline |
| Continue | Continue | Acquired by Cursor (June 2026); hosted product shut down, open-source codebase frozen at v2.0.0; VS Code + JetBrains |
| Augment Code | Augment | World-class context engine; enterprise-grade; strong SWE-bench scores |
| Cody | Sourcegraph | Deep codebase indexing; cross-repository context; best for large codebases |
| Supermaven | Supermaven | Ultra-fast AI completion; large context windows |
| Blackbox | Blackbox | AI code completion + chat; multi-language support |
| Snyk Code | Snyk | AI-driven security scanning; real-time vulnerability detection in-IDE |
| Qodo | Qodo | AI-generated unit tests; code quality + PR review focus |
| MarsCode | ByteDance | AI coding assistant and online IDE; companion to Trae |
| Replit AI | Replit | AI suite inside Replit's cloud IDE; beginner-friendly; includes Ghostwriter |
| CodeBuddy | Tencent | MCP-compatible coding assistant extension; ⚠️ not available in the USA |
| Cortex Code | Snowflake | MCP-compatible AI code assistant extension |
| Kode | ShareAI | AI developer assistant and workspace integration |
| Apertis | Stima | One API key works across all major coding agents |
| Corust | Corust | Your seasoned Rust co-pilot: production-grade code generation, zero hallucinations on Rust idioms, and tools built for real Rustaceans. ⚠️ Site unreachable: corust.ai has no DNS A/AAAA record (verified Sep 2026); link removed |
| Toprank | nowork-studio | Open-source Claude Code plugin for SEO, Google Ads, content writing, and CMS optimization workflows |
| WozCode | WozCode | A Claude Code plugin that supercharges performance, cost, and speed |
| Claude Code Skills 中文精选集 | 老实人实验室 | Chinese curated directory of Claude Code Skills, Agents, Plugins, and workflows with 140+ resources |
| Oh My codeX | Yeachan-Heo | OmX: Your codex is not alone. Add hooks, agent teams, HUDs, and so much more |
| Bob | IBM | Your AI-Powered Development Partner |
| Firebender | Firebender | The first Android-native coding agent that writes features, tests them in the emulator, and fixes issues automatically |
| Feather Wand | NaveenKumar Namachivayam | Supercharge your performance testing workflow with AI-driven capabilities built natively into JMeter.; GitHub |
| SF Pi | Salesforce | Opinionated Salesforce extensions for the Pi coding agent: focused lifecycle tools, Salesforce-aware status and safety surfaces, Agent Script authoring, and one Manager for package settings and extension enablement.; GitHub |
| Blume | Blume | Blume helps you build an agentic-native codebase with consistent and correct context. Fewer mistakes, fewer headaches. |
| Codex++ | bennett | Codex++ tweak system for the Codex desktop app; GitHub |
| Ponytail | Dietrich Gebert | Open-source (MIT) agent plugin/skill that makes coding agents "think like the laziest senior dev": a seven-rung YAGNI ladder (reuse, stdlib, native platform, one line) applied before writing code, benchmarked at ~54% less code and ~20% lower cost on real Claude Code sessions; installs as a plugin for Claude Code, Codex, Copilot CLI, Cursor, Pi, and ~20 agents; GitHub |
| AICode | AI Sovereign Labs | AICOde is an AI coding assistant, as a VS Code extension: Spec-driven ; human validation gate before code generation ; MCP/skills support ; BYOK Azure/OpenAI ; 5D local index ; built-in QA harness. Specialized in quality code generation using a 5 step workflow: Ideate → Specify → Refine → Code → Verify. |
💻 Terminal & CLI Agents
AI coding agents that live in your terminal or command line.
| Tool | Company | Notes |
|---|---|---|
| Claude Code | Anthropic | Terminal-native agentic coding; top-rated for complex reasoning & large refactors |
| Codex CLI | OpenAI | Re-emerged as agent-first tool; runs against real repos from the CLI |
| Amp | Sourcegraph | Multi-model coding agent available through a CLI and VS Code, with shareable threads, MCP, custom skills, and an SDK |
| Aider | Aider | Git-native terminal pair programmer; 39K GitHub stars; auto-commits changes |
| Gemini CLI | Terminal agent for local repo tasks; lightweight; no UI overhead | |
| Goose | Block | Open-source AI agent framework; fully local; write/execute/debug from CLI |
| Amazon Q CLI | AWS | CLI agent component of Amazon Q; AWS-integrated terminal tasks |
| Letta Code | Letta | Memory-first terminal coding agent; stateful across sessions; Docs |
| Qwen CLI | Alibaba | High-performance CLI tool for Qwen models; specialized in code generation |
| Codebuff | Codebuff | Terminal-native AI coding agent with deep codebase awareness; GitHub |
| OpenClaw | OpenClaw | CLI-based autonomous agent with MCP skill support |
| Command Code | Command Code | Stop fixing sloppy AI code. Command Code continuously learns your coding taste. Powered by taste-1 applied meta neuro-symbolic AI. |
| Crush | Charmbracelet | CLI tool for MCP-based coding workflows |
| iFlow CLI | iFlow | Terminal AI agent for workflow automation |
| Kiro CLI | AWS | Command-line agent interface with MCP capabilities |
| lucinate | lucinate-ai | Terminal-native TUI chat client for AI agents; connects to OpenClaw, Hermes, and any OpenAI-compatible endpoint; Homebrew tap available |
| MCPJam | MCPJam | CLI agent optimized for Model Context Protocol integration |
| Mistral Vibe | Mistral | Terminal agent leveraging Mistral models |
| Mux | Coder | Advanced terminal AI assistant and multiplexer |
| Zencoder | Zencoder | Specialized terminal-based AI coding assistant |
| Neovate | Neovate | Next-gen terminal and Neovim integrated agent |
| AdaL | Sylph | Autonomous developer agent for the CLI |
| Cline Kanban | Cline | Cline Kanban works with the agents you're already using: Claude Code, Codex, and Cline-compatible agents, with more to come. |
| Parallel Code | Parallel Code | Parallel Code is a desktop app that gives every AI coding agent its own git branch and worktree, automatically. |
| Pi | badlogic | Pi is a minimal terminal coding harness. Adapt pi to your workflows, not the other way around. |
| Autohand ACP | Autohand | ACP-based terminal coding agent with autonomous capabilities |
| Crow CLI | Crow | Two-layer system: ACP agent (crow-cli) that does the thinking, and MCP toolserver (crow-mcp) that does the doing |
| Fast Agent | EvalState | Code, Build and Evaluate agents: excellent Model and Skills/MCP/ACP Support |
| Minion Code | Femto | Minion's implementation of Claude Code |
| Compass Nova CLI | Compass | Compass Nova CLI |
| Pi ACP | svkozak | ACP adapter for pi coding agent |
| Stakpak Agent | Stakpak | Ship your code, on autopilot. Open source agent that lives on your machines 24/7 and keeps your apps running |
| OpenCode | Anomaly | OpenCode is an open source agent that helps you write and run code with any AI model. It's available as a terminal-based interface, desktop app, or IDE extension. Free models included or connect any model from any provider, including Claude, GPT, Gemini and more. |
| Antigravity CLI | The terminal-first surface to interact with Antigravity agents. Stay in your flow without context switching. | |
| Nanocoder | Nano Collective | Local-first, open-source terminal coding agent built by a community collective. Bring your own model (Ollama, OpenRouter, or any OpenAI-compatible API) and keep your code on your machine, with native tool calling, an XML fallback, MCP server support, and file-based custom commands and tools.; GitHub |
| AI Prompt Architect | AI Prompt Architect | Production-grade prompt engineering platform. Includes a 33-command CLI & MCP Server integration that offers a 50% API credit discount for AI agents, plus the STCO framework and multi-model evaluation. |
| AI Badger | PVR Labs | AI Badger is a local-first CLI that maps your codebase and extracts focused file context for AI chats and coding agents. It works with ChatGPT, Claude, Codex, Gemini, and other assistants without cloud indexing, API keys, or vendor lock-in.; GitHub |
| AgentBox | Marco D'Alia | Run multiple coding agents (Claude Code, Codex, OpenCode) in parallel, each teleported into its own sandboxed VM: local Docker, self-hosted, or cloud (Hetzner/Daytona/E2B). Sub-second checkpoints, per-box browser/VS Code/persistent shells, native macOS menu-bar app, git credentials kept on the host. MIT.; GitHub |
| Oh My Pi | Can Bölük | ⌥ AI Coding agent for the terminal: hash-anchored edits, optimized tool harness, LSP, Python, browser, subagents, and more. |
| Grok Build | X | A powerful coding agent for complex work. |
| BitFun | GCWing | Open-source coding agent with an interactive terminal UI and desktop app. It can plan, edit, test, and commit inside real Git repositories, with protected tool calls requiring approval by default.; GitHub |
| DeepSeek Harness | DeepSeek | DeepSeek Harness (dsh) is an open-source agent harness developed by DeepSeek AI.; GitHub |
| Atomic Agent | AtomicBot | Local-first CLI and TUI coding agent that runs open-weight models entirely on your machine through a llama.cpp fork, so no account or API key is required. Ships 56 built-in tools (browser, filesystem, git, memory, vision), MCP support, and a five-layer local memory system.; GitHub |
| Kane CLI | TestMU AI | End-to-end flows on your local browser, powered by natural language. Instant validation, deep bug discovery, and production-ready automation that elevates release confidence. |
| Caveman Code | Julius Brussee | Same model. Same task. ~2× fewer tokens than Codex. 20+ providers · plan mode · autopilot loop · MIT.; GitHub |
| Sillage | MarlBurroW (open source) | Self-hosted, MIT-licensed, mobile-first web UI that drives the native Claude Code and Codex CLIs on your own machine (the official agent harnesses, without a terminal). Sessions that outlive the client, full-text search over every conversation, an IDE panel (file explorer, editor, diffs, terminal), and an installable PWA. Single Docker container.; GitHub |
| MathCode | Math-AI | MathCode: A Frontier Mathematical Coding Agent; GitHub |
| fx.sh | Vercel | fx is a coding agent harness and CLI written in Zig, optimized for research and embeddability as part of larger systems.; GitHub |
| Zero | Gitlawb | Zero is an AI coding agent for your local terminal. It can inspect a repository, edit files, run commands, use browser/terminal helpers, and keep durable local sessions while you choose the model and the permission level.; GitHub |
| Headlong | Laude Institute | An open source agent microharness featuring persistent agency and recursive LLMs. Of bash, by bash, for bash; it's shells all the way down.; GitHub |
| OB-1 | Overbrilliant | Terminal coding agent whose default path needs no account, no API key and no card: npm i -g @overbrilliant/ob1, run ob1, and an embedded router inside the CLI process answers the first message through keyless free providers using a signed bundled free-model catalog, with failover to your own keys (OpenRouter, OpenAI, Gemini, Groq) or local runtimes (Ollama, LM Studio, llama.cpp, vLLM, any OpenAI-compatible endpoint). It also ships persistent SQLite project memory with an inspectable fact/relationship graph, a repo map, MCP support, and OS-level sandboxing (Seatbelt on macOS, bubblewrap on Linux). To be straight about the trade-off: the keyless tier is a bootstrap path with variable model quality and shared limits, not a frontier-model tier. Apache-2.0; also installable via brew install overbrilliant/tap/ob1.; GitHub |
| XIRP | Spotify | Xirp connects to your services, ownership, docs, and architectural decisions, so every AI coding session starts with real context, not guesswork. |
| Kolkrabbi | onembyte (Franco Michetti) | Go single-binary terminal coding agent (kolk), Apache-2.0. One effort dial (low/medium/high/max) selects which model runs the turn rather than how many thinking tokens one model spends. Runs any OpenRouter model, a local model on Ollama, vLLM or any OpenAI-compatible URL, or your own Claude and ChatGPT subscriptions by driving each vendor's CLI as a subprocess without holding a credential. Opt-in OS sandbox (Seatbelt on macOS, Landlock on Linux). Every call is logged locally with cost, latency and your own ratings; no telemetry.; GitHub |
| FrontierAgent | Apodex AI | Open-source (Apache-2.0) agent runtime and terminal TUI (frontier-agent) for long-horizon research and file-based work: ReAct single-agent mode and an Agent Team mode where a coordinator keeps a live task board and delegates to parallel sub-agents; task-scoped sandbox (/inputs, /workspace, /outputs), Plan/Activity/Files/Diff sidebar, approvals, mid-run steering, session resume, and any OpenAI-compatible endpoint (Apodex API or self-hosted SGLang) |
| YYLO | yylo-dev | Open-source command-line orchestrator for coding agents, repeatable workflows, and receipt-backed repository changes (MIT; npm @yylo/cli). Every task runs in a dedicated branch/worktree with typed task, validation, merge, and release-readiness boundaries; the merge queue owns risk-based review, and runs retain declared receipt hashes and terminal manifests. Orchestrates Pi and Codex coding subagents.; GitHub |
| pi-agent-ide | Aleksandr Shpuntenko | Precise editing tooling, benchmaxxed and forged in reallife daily coding.; GitHub |
| MiMo Code | Xiaomi | Open-source (MIT) terminal-native coding agent from Xiaomi with build/plan/compose agents, SQLite FTS5-backed persistent project memory, automatic checkpoints and context reconstruction, mimo serve/attach for remote sessions, and connections to Xiaomi MiMo, Codex OAuth, imported Claude Code auth, or any OpenAI-compatible provider. npm install -g @mimo-ai/cli; also the engine behind the MiMo Desktop beta.; GitHub |
| molt | Solvyx / Tyler Skelton | Coding agent (CLI/TUI + desktop) that won't say done on a false claim. Verification runs against on-disk checks (.molt/done.yml); receipts for accepts and refusals. OpenAI-compatible or Anthropic. Install from GitHub releases or build from source. Studio: https://solvyx.xyz/work/molt |
| Prime Agent | Prime Intellect | Open-source (MIT) terminal coding and research agent built on Recursive Language Models: a persistent Python REPL is the model's only built-in tool, so file ops, shell commands, context management and real child agents (rlm.spawn(...)) all run as code. Daemon-backed sessions survive terminal disconnects (prime-agent attach), /goal, /heartbeat, schedules and a budgeted /autonomous mode keep long tasks moving, and /refine persists evidence-backed lessons as supplemental prompts, memories and skills (Continual Harness, arXiv 2608.23552). One-command installer from app.primeintellect.ai; /login with a subscription or API key. |
| CodeAF | AgentField | Open-source (Apache-2.0) Go terminal coding harness built for open-weight models (early preview): one window lists every project and conversation, turns requests into parallel tasks on isolated branches, auto-picks a worker/planner/checker model crew per task with spend limits, and merges only work that passes its check; connects OpenRouter, DeepSeek, GLM, Kimi, MiniMax, Qwen, Ollama, Codex via a ChatGPT plan or any OpenAI-compatible endpoint; GitHub |
| Herdr | Herdr | Open-source (Apache-2.0) agent multiplexer and "runtime your coding agents live on": one Rust binary (no Electron) that runs inside your existing terminal and owns the terminals of Claude Code, Codex, Cursor, OpenCode, Grok and other CLI agents without wrapping them; a background server keeps panes running when you detach or lose SSH and restores layouts after restarts, every pane is marked working, blocked or idle, local and saved SSH machines share one window, tmux-style prefix keys plus mouse splits, a plugin marketplace, and a CLI/socket API plus agent skill so agents can spawn panes, prompt each other and wait on blocked peers; install via the one-line script at https://herdr.dev/install.sh, Homebrew (brew install herdr) or mise; GitHub |
| Agent Deck | Ashesh Goplani | Open-source (MIT) Go terminal session manager for AI coding agents (macOS, Linux, WSL): one TUI shows every Claude Code, Codex, Gemini, OpenCode and Pi session as running, waiting or idle; fork sessions with inherited history, per-project MCP Manager and Skills Manager toggles, Git worktree and sandboxed container sessions, an agent-deck web UI, and a "conductor" session that watches the fleet and answers or escalates over Telegram, Slack, Discord or Mattermost; install with brew install asheshgoplani/tap/agent-deck, the install script or go install |
💻 AI-Native Terminals
Full terminal emulators built from the ground up with AI capabilities integrated.
| Tool | Company | Notes |
|---|---|---|
| Warp | Warp | The wildly popular rust-based terminal with "Warp AI" built directly into the command line interface |
| Wave | Wave | An open-source, modern AI-native terminal alternative |
| TermRover | TermRover | Native SSH and Mosh terminal for iPhone and Android, optimized for tmux with mobile-friendly tabs, prefix actions, and scrolling |
| uniTerm | ys-ll | Open-source (Apache-2.0) all-in-one desktop terminal (Go + Wails v3, Vue 3, xterm.js) for Windows, macOS, Linux and Android covering 30+ protocols (SSH/Telnet/Mosh/serial, SFTP/SCP/FTP/SMB/WebDAV/S3, RDP/VNC/SPICE/X11, MySQL/PostgreSQL/Redis/MongoDB and more, Kubernetes/Docker/Podman) with a built-in AI agent that plans and runs multi-turn shell commands in the active tab using Anthropic- or OpenAI-compatible models (four approval modes from bypass to confirm-all), natural-language database queries, reusable skills, and an MCP server that lets Claude Code, Codex or Gemini CLI run commands and transfer files on saved connections without exposing credentials; GitHub |
| cmux | Manaflow | Open-source (GPL-3.0-or-later) Ghostty-based macOS terminal built for running many AI coding agents at once: vertical tabs show git branch, linked PR status, working directory, listening ports and the latest agent notification; panes light up with notification rings when Claude Code, Codex or OpenCode need attention; scriptable in-app browser split, cmux ssh remote workspaces, cmux claude-teams for Claude Code teammate mode as native splits, session restore and a CLI/socket API; free DMG or Homebrew install; GitHub |
| Corgi | Tal Shiri (tals) | Open-source (MIT) agent-aware terminal multiplexer powered by Ghostty and inspired by Herdr (one-line install.sh installer; macOS Apple Silicon, Linux amd64, WSL): detects Claude Code, Codex, Pi and OpenCode in any pane and shows each agent's status (working, needs you, done), activity summaries, subagents, context/token/rate-limit usage, multi-account launch and resume after crash or reboot; sessions keep running when detached like tmux, F1 jumps to the next unread agent, and a corgi CLI (list, pane send, pane read, --json, --skill) lets scripts and agents drive panes; GitHub |
🤖 Autonomous & Async Agents
Agents that operate independently on tasks, often outside your local editor.
| Tool | Company | Notes |
|---|---|---|
| Jules | Async GitHub-integrated agent; clones repo into GCP VM; works while you code; powered by Gemini 2.5 Pro | |
| OpenHands | All Hands | Open-source autonomous coding agent; 95K+ GitHub stars |
| SWE-agent | Princeton NLP | Open-source agent for autonomously solving GitHub issues |
| Sweep | Sweep | Acts as an autonomous junior developer. Branches, writes code, and opens PRs from issues |
| Ellipsis | Ellipsis | AI agent that reviews pull requests and converts comments directly into runnable code commits |
| Genie | Cosine | One of the highest performing autonomous models strictly evaluated against the SWE-bench benchmarks |
| Open-yak | Open-yak | Open-source desktop AI agent with 130+ skills, 46 MCP connectors, and IM gateway |
| DeepAgents JS | LangChain | TypeScript package for creating Deep Agents: implements deep agent capabilities in a general-purpose way |
| Orkas | Orkas-AI | Open-source, local-first desktop workspace where a Commander decomposes goals and coordinates specialist AI agents in parallel or sequence.; GitHub |
| Kody | Kent C. Dodds | Your assistant's home: the memory, keys, code, and automations your AI agent keeps, portable across every MCP host. Built on Cloudflare Workers.; GitHub |
| Omnigent | Omnigent | Omnigent is an open-source AI agent framework and meta-harness: orchestrate Claude Code, Codex, Cursor, Pi, and custom agents: swap harnesses without rewriting, enforce policies and sandboxing, and collaborate in real time from any device.; GitHub |
| Manus | Monica | Manus is an autonomous AI agent that can plan, code, and execute complex tasks end-to-end in the browser. |
| LoopTroop | LoopTroop AI | Local, open-source GUI for running multi-step AI coding tickets across projects. Uses an LLM council for planning, atomic Beads in isolated git worktrees, and time-boxed retry loops for execution.; GitHub |
| Muse | Meta | Personal AI agent that pursues goals and everyday tasks using a secure persistent computer, connected apps, approvals, and proactive suggestions |
| OtoDock | OtoDock | Fair-source, self-hosted agentic company OS for multi-tenant teams that lets Claude Code, Codex, and local models run collaborative agents with memory, tools, delegation, schedules, phone access, and isolated workspaces |
| firstmate | kunchenguid | Open-source (MIT) "agent distro" for running a crew of coding agents: you talk to one first-mate agent (Claude Code, Codex, OpenCode, Pi, Grok, or Cursor Agent CLI) and it spawns autonomous crewmates in tmux/zellij/cmux/Orca sessions, gives each a clean git worktree, supervises them with a zero-token bash watcher, and hands back finished PRs, local merges, or investigation reports; optional Relay answers mentions on X/Discord; GitHub |
| OpenResearch | alphaXiv | Open-source (MIT) local-first workspace (orx up) that turns Claude Code, Codex, OpenCode, or Cursor into research agents: parallel agent sessions in isolated git worktrees, a git-native experiment tree with immutable run archives, autonomous "autoresearch" propose → run → inspect loops, and the same snapshot runnable locally, over SSH, or on Slurm/Kubernetes/Ray/Modal/HF Jobs; GitHub |
| Hermes Agent | Nous Research | Self-improving agent with a built-in learning loop that creates and refines its own skills, searches past sessions, and builds a persistent user model. Runs as a terminal TUI or via Telegram, Discord, Slack, and more, on local machines, Docker, SSH, or serverless backends like Modal and Daytona |
| Coder | Coder | Open-source (AGPL-3.0) self-hosted platform for cloud development environments and AI coding agents: workspaces defined in Terraform (Kubernetes, Docker, EC2), connected over WireGuard, auto-shutdown when idle; Coder Agents runs a native coding agent whose loop executes in the control plane on your infrastructure with no LLM keys in workspaces, bring-any-model (Anthropic, OpenAI, Google, Bedrock, self-hosted), user identity on every action, plus an AI Gateway for auth, audit, and cost controls; registry modules run Claude Code, Codex, and OpenCode isolated in workspaces; GitHub |
| AX | AX is a high-throughput, declarative orchestrator to run billions of autonomous agent workloads in a cluster. It runs on top of Agent Substrate for sandboxed execution and is built to run billions of tasks per cluster. If you have used Kubernetes, ax will feel similar. | |
| GitHub Agentic Workflows | GitHub | Open-source (MIT) gh extension that defines AI-powered repository automation in Markdown with YAML frontmatter and compiles it into GitHub Actions; runs Copilot, Claude Code, Codex, Gemini, or Pi agents securely with MCP tools, safe-outputs, and network sandboxing |
| Strand Agents | Strand | Build an agent harness and control it end-to-end. Open-source SDK for production AI agents in Python & TypeScript - any model, any cloud. |
| Paperclip | Paperclip Labs | Open-source (MIT) Node.js server + React UI that orchestrates a team of AI agents like a company rather than a pile of terminals: org chart with roles and reporting lines, goal-aligned ticketed tasks, scheduled heartbeats, per-agent monthly budgets with atomic task checkout, approval gates with revisioned/rollback-able config, full tool-call tracing and audit log, multi-organization isolation and a mobile UI; bring any agent (OpenClaw, Claude Code, Codex, Cursor, bash, HTTP) or run them in e2b/Cloudflare/Daytona/Modal sandboxes; GitHub |
| OpenSRE | Tracer | Open-source (Apache-2.0) framework for building AI SRE agents that investigate production incidents on your own infrastructure: the opensre CLI (REPL or headless opensre ask "why is checkout-api slow?") correlates logs, metrics, traces, deploys and runbooks across 60+ integrations (Grafana, Datadog, Honeycomb, Kubernetes, AWS, PagerDuty, Opsgenie, Slack and more), posts root-cause summaries back to Slack or PagerDuty, and ships local and cloud-backed end-to-end incident test suites as a training/evaluation environment for AI SRE agents; public alpha with near-daily releases, self-hostable FastAPI deployment or the hosted OpenSRE Cloud; GitHub |
| SuperPlane | SuperPlane | Open-source (Apache-2.0) "software factory for one-shot engineering": scans your repos and backlog, scores tickets by automation confidence, dispatches the coding agent you configure on managed or self-hosted runners, then gates the branch/PR on CI status, repository checks and review feedback, returning failures to the agent with output attached before a human reviews; git-backed workflow apps with approvals and durable execution across GitHub, GitLab, CI/CD and incident tools; self-host the engine or use SuperPlane Cloud.; GitHub |
| Docker Agent | Docker | Open-source (Apache-2.0) agent builder and runtime from Docker Engineering: define single agents or multi-agent teams in declarative YAML with built-in think/todo/memory tools, pluggable RAG and any MCP server (local, remote or Docker MCP Toolkit), run them via the docker agent CLI plugin (pre-installed in Docker Desktop 4.63+, or brew install docker-agent) with a TUI, MCP mode and evals, and push/pull agents as OCI artifacts; provider-agnostic across OpenAI, Anthropic, Gemini, Bedrock, Mistral, xAI and local Docker Model Runner models; GitHub |
| Orbi | orbi-build (Lawrence Liu) | Open-source (AGPL-3.0) agent that turns a labeled GitHub issue into a merged PR: one AI run writes the change, a separate AI run reviews it against the issue's acceptance criteria, then merges the approved commit on its own (branch protection can require your approval). Release issues tag and publish versions. Self-host free with your own model key, or use Orbi Cloud (first 3 merged PRs free, then $29/month, model tokens included).; GitHub |
| OpenHuman | Tiny Humans | Open-source (GPL-3.0) agent harness written in Rust, shipped as a desktop app (macOS/Windows/Linux), terminal app, headless server and embeddable Rust library: agent loop with shell/coder tools, MCP servers and skills, built-in memory, approval gates and sandboxed execution (OS jail or Docker), BYOK or local models (Ollama, LM Studio, MLX), and a single-process core light enough to run hundreds of agents on a small VPS; GitHub |
| Reason | Reason Machines | Autonomous software engineer for GitHub: cloud agents turn assigned issues into evidence-backed PRs with screen recordings, logs and review-ready diffs; web app, macOS desktop and CLI, HTTP API and MCP server; Reason Agent 1.1 claims a higher Coding Agent Index than Codex at 22% lower cost per solved task; GitHub |
| Rakazo | Elie Steinbock (elie222) | Open-source (Apache-2.0), self-hosted agent platform for persistent AI "teammates" (beta) from the creator of Inbox Zero: each bot keeps its own conversations, memory, routines and history, works on shared Team Computers or isolated Private computers with browser, terminal, file and graphical desktop access, and can delegate to peer bots or short-lived subagents; bring your own model credentials (via Pi) and sandbox provider (Docker, E2B, Daytona, CreateOS, Box or a trusted local machine), with app integrations through Composio, Pipedream Connect, remote MCP, OpenAPI and Treg; voice mode, web UI, Electron desktop app and iPhone/Android apps; one-command Docker Compose install; positioned as a self-hosted alternative to Grok Bot, OpenClaw and Hermes Agent; GitHub |
| Open-CUAK | Aident AI | Open-source, self-hostable kit for running and managing computer-use automation agents, starting with browsers, as a free alternative to OpenAI Operator. Runs locally via Homebrew or Docker.; GitHub |
| Omnara | Omnara (YC S25) | Open-source (Apache-2.0) managed agent platform and self-hostable alternative to Claude Managed Agents and OpenAI's Agents API: define an agent in YAML, launch it via REST API, TypeScript SDK (@omnara/sdk) or npx omnara CLI, and Omnara runs the durable agent loop (state in Postgres, survives restarts, idle agents cost nothing) with sandboxes or your own laptop/VM as machines, any model (OpenAI, Anthropic, OpenRouter, self-hosted), secrets, permissions/RBAC and event streaming; hosted Omnara Cloud also available; GitHub |
| Scion | Google Cloud Platform | Open-source (Apache-2.0), experimental orchestration platform for teams of AI agents: runs Claude Code, Antigravity, Codex and OpenCode as independent containerized agents with per-agent config and credentials, shared workspaces or separate Git worktrees, roles, delegation and messaging; scion CLI plus Workstation web UI, scaling to hosted Hubs on Docker, Podman, Apple Container, Kubernetes or Cloud Run; pre-1.0 and not an officially supported Google product; GitHub |
| Cindy | XD Inc. | Open-source (Apache-2.0) desktop and mobile AI agent that runs locally on your files and logged-in apps, wrapping the Claude Code, Codex and Pi harnesses (native harness in the works) so models and harnesses can be mixed or switched mid-task while memory, skills, MCP tools and the workspace stay continuous; one task can be planned, executed in parallel and reviewed by agents on different harness and model combos, and Cindy can drive the browser, computer and Android phone, connect over SSH, and take work from Slack, Feishu, GitHub, Jira and schedules; Electron app for macOS, Windows and Linux plus iOS and Android companions; free with your own API keys, existing Claude Code / Codex / Pi coding plans or local models, with an optional paid managed model service; GitHub |
| Zozo HQ | Roundhouse Labs Inc. | Zozo HQ orchestrates dependency-ready issues into pull requests using isolated worktrees, human clarification and a recovery journal, with Claude Code, Codex or OpenCode via OpenRouter (a usable runtime must be installed), on GitHub or matched Jira Cloud + Bitbucket Cloud workflows. The Local Stack is free and hosting starts at US$40/month before tax, with model access billed separately; submitted by founder Ryan Graham with AI assistance. |
| Agensh | Microsoft | Open-source (MIT) research harness from Microsoft that lets a team of coding agents (scaled to 1,024 in the paper) self-organize without a central orchestrator: agents claim sub-tasks, share findings, verify results and merge progress asynchronously through a shared Gitea workspace, Mattermost messaging and shared context. The released single-node Docker Compose setup runs each worker on GitHub Copilot ACP sessions and targets Linux.; Project page |
| OpenBot | Nightly Labs | Source-available (PolyForm Noncommercial 1.0.0) local-first desktop workspace for persistent AI teammates: each agent gets its own workspace, conversation and FIFO task queue backed by an independent Codex, Claude Code, Grok CLI, OpenCode, Gemini, Cursor CLI or Cline CLI session (via ACP), with custom OpenAI-compatible endpoints and local Ollama/LM Studio detection; agents message each other, hand off files, share channel chats with explicit delegation, drive a persistent embedded browser and optional Computer Use, and get automatic context compaction; Electron app for macOS, Windows and Linux plus an iPhone app, a headless Linux server and Docker image; development preview (agents run with full access after first-launch consent); free for noncommercial use, commercial use needs a separate license; GitHub |
| AutoSwarm | HKUDS | Open-source (MIT) multi-agent coordinator from HKU Data Intelligence Lab (backed by the arXiv paper "AutoSwarm: Automating Agent Swarms") that installs as a skill/plug-in into the agent you already use and turns your locally installed agents (Codex, Claude Code, OpenCode, Cursor, GitHub Copilot CLI, Muse Code, CodeBuddy, Hermes, nanobot, DeepSeek Harness, Grok Build) into one team: it picks agents by capability, shares context and memory across them, runs subtasks in parallel and verifies and synthesizes results; Python autoswarm CLI with setup --host and doctor --probe, reuses native credentials, works on Windows, macOS and Linux |
🌐 Browser-Based & App Builders
AI tools that generate full apps or UIs from natural language, no local setup required.
| Tool | Company | Notes |
|---|---|---|
| Bolt.new | StackBlitz | Prompt → full-stack app in browser; React/Vue/Node; viral vibe-coding tool |
| v0 | Vercel | Natural language → React + Tailwind UI components; Vercel ecosystem |
| Lovable | Lovable | AI web app builder; production-ready full-stack from prompts |
| Replit | Replit | Cloud IDE + AI builder; deploy from browser; great for beginners |
| PlayCode | PlayCode | 15+ AI models in-browser; builds complete websites without coding |
| Emergent.sh | Emergent | Agentic vibe-coding platform for building and deploying full-stack apps from natural language |
| Bitrig | Bitrig | Describe what you want to build, and Bitrig turns it into real Swift code you can ship to the App Store. |
| Floot | Floot | Floot turns Claude or ChatGPT into a full app builder - backend, database, and hosting included. No git, no terminal, no build credits. |
| Dyad | Dyad | Local, open-source AI app builder with BYO model support, no sign-up, and no vendor lock-in; GitHub |
| Onlook | Onlook | Open-source visual AI editor for building, styling, and editing React/Tailwind apps directly in code; GitHub |
| Base44 | Base44 | Natural-language platform for building and hosting full-stack apps, websites, and AI agents with managed data, auth, integrations, and domains |
| Ploy | Ploy | AI marketing platform that builds and hosts on-brand websites as production-grade Astro code with GitHub sync, then runs "Ploybook" agents that continuously fix technical SEO, write content, identify visitors, and run ads experiments; free tier with 800 credits/day |
| Rork | Rork | Prompt-to-mobile-app builder for iOS, Android, and web with live simulation and App Store/TestFlight publishing |
| Cloudflare OS | Cloudflare | Open-source (Apache-2.0) agent workspace built on Cloudflare Workers that Cloudflare runs internally: an agent chat UI grounded in company context and skills, sandboxed "Gadgets" (personal apps agents build, each instance in its own Dynamic Worker Facet) that can be shared and modified, and capability-based "Gatekeepers" that broker OAuth access to services like Google Workspace with logging and human-in-the-loop approval; deploy to your own Cloudflare account or workerd, managed deployments on waitlist.; GitHub |
| Modaal | Modaal | Mac app that builds native iOS and Android apps in SwiftUI and Jetpack Compose with the AI coding agent you already pay for, from a spec you approve before code is written. |
| OpenPencil | ZSeven-W | Open-source (MIT) AI-native vector design tool (desktop app for macOS/Windows/Linux plus op CLI): prompt-to-UI on a live canvas where concurrent "agent teams" build page sections in parallel, Git-friendly JSON .op design files, and code export to React + Tailwind, HTML/CSS, Vue, Svelte, Flutter, SwiftUI, Jetpack Compose and React Native; a built-in MCP server installs into Claude Code, Codex, OpenCode, Kiro or Copilot CLI so terminal agents can read and edit designs; GitHub |
🛡️ AI Code Review & Security
Tools focused on reviewing, securing, and validating code, not generating it.
| Tool | Company | Notes |
|---|---|---|
| CodeRabbit | CodeRabbit | AI-powered PR reviews; line-by-line feedback; integrates with GitHub/GitLab |
| Qodo | Qodo | PR validation, test generation, merge readiness checks |
| Snyk Code | Snyk | SAST security scanning; real-time vulnerability detection; DevSecOps integration |
| cubic | Cubic | AI code reviews for complex codebases |
| Kodus | Kodus | Open source AI code review tool that helps engineering teams review pull requests with repository context, custom rules, and BYOK support. It integrates with GitHub, GitLab, Bitbucket, and Azure DevOps.; GitHub |
| Gito | Vitalii Stepanenko | Open-source AI code reviewer that runs in GitHub Actions or locally. Model-agnostic (works with any LLM) and reports findings to GitHub, Jira, or Linear.; GitHub |
| Bubo | MountainOwl | I maintain Bubo, a self-hosted AI reviewer for GitHub PRs and GitLab MRs. I built it to run CLI-driven models, post evidence-backed inline findings or LGTM, and learn from maintainer feedback to reduce repository-specific noise.; GitHub |
| heygrc | ISMS Copilot / Better ISMS | GitHub App that reviews every pull request against compliance frameworks (ISO 27001, SOC 2, GDPR, EU AI Act, and more), cites the control clause, and says what to fix. Public repositories always free. Install: https://github.com/apps/heygrc |
| Mydentify AI Crawler Access Checker | Mydentify / Timothy Allard | A free browser-based checker that tests whether AI crawlers can access a site by inspecting robots.txt, page-level directives, response headers, and user-agent behavior. It reports observable access signals and their limits; it does not claim to measure crawler indexing or guarantee AI visibility.; GitHub |
| Vercel Agent | Vercel | Production-aware agent for investigations and PR reviews; validates proposed fixes in secure sandboxes against real builds, tests, and linters |
| Graphite | Graphite | AI code review platform combined with stacked pull requests, merge queues, and engineering workflow analytics |
| Bito | Bito | Context-aware AI code review for GitHub, GitLab, and Bitbucket with cross-repo impact analysis, custom rules, and one-click fixes |
| security-audit | Cloudflare | Open-source (MIT) coding-agent skill that seeded Cloudflare's fleet-wide vulnerability discovery harness: runs a six-phase audit (reconnaissance with coverage ledger → coverage-led hunting by isolated hunter agents → adversarial validation → schema-checked findings.json → independent record verification → target-neutral reports) with per-domain attack-class playbooks for web/auth, supply chain, cloud, LLM prompt injection, memory safety, and more; additive across runs; GitHub |
| Open Code Review | Alibaba | Open-source (Apache-2.0) AI code review CLI (ocr) incubated from Alibaba's internal reviewer; hybrid deterministic rules + LLM agent with line-level comments, ocr scan for whole-file audits, and integrations for Claude Code, Codex, and Cursor; GitHub |
| SkillSpector | NVIDIA | Open-source (Apache-2.0) security scanner for AI agent skills (Claude Code, Codex, Gemini CLI, MCP): point it at a Git repo, URL, zip, directory or single SKILL.md and it runs fast static checks (AST walk, taint tracking, YARA, OSV.dev dependency lookups) across 70+ vulnerability patterns in 17 categories (prompt injection, data exfiltration, privilege escalation, supply chain, MCP tool poisoning) with optional LLM semantic review to cut false positives; 0–100 risk score, baselines, and terminal/JSON/Markdown/SARIF output for CI gates; also runs as an MCP server and Pi/OpenCode extension; GitHub |
| Strix | Strix | Open-source (Apache-2.0) autonomous AI penetration testing agents that run your app dynamically and validate findings with working proofs-of-concept instead of static-analysis guesses: multi-agent recon/exploitation/validation with an HTTP proxy, browser automation, shells and a Python runtime in a Docker sandbox, auto-fix patches and pentest reports; GitHub Actions PR scanning, npx skills add usestrix/strix for Claude Code/Cursor/Codex, and a managed platform at app.strix.ai.; GitHub |
| ADR | Uber | Open-source (Apache-2.0) Agentic AI Detection and Response system deployed at Uber (MLSys 2026 paper): ADR Discovery inventories AI apps, CLI agents, IDE extensions, local model runtimes and MCP servers on endpoints, ADR Sensor normalizes session telemetry from Claude Code, Cursor, Codex, GitHub Copilot CLI, Gemini CLI, opencode and others, and a dual-agent detector is evaluated on ADR-Bench (300+ tasks, 134 MCP servers, 17 attack techniques); GitHub |
| Codex Security | OpenAI | Open-source (Apache-2.0) CLI and TypeScript SDK (npx @openai/codex-security scan) for finding, validating and fixing security vulnerabilities: scans whole repos, selected paths or Git diffs (deep mode runs parallel discovery workers), validates candidate findings, generates and verifies patches, saves threat models, drafts SECURITY.md policies, checks for exposed credentials, imports GitHub code scanning alerts and exports SARIF/JSON/CSV or publishes to Linear; runs in CI and containers with ChatGPT login, an OpenAI API key, Amazon Bedrock, OpenRouter or Fireworks AI; GitHub |
| Plannotator | backnotprop | Open-source (Apache-2.0/MIT) local, browser-based review surface for coding agents: when Claude Code, Codex, Copilot CLI, Gemini CLI, OpenCode, Kiro, Droid, Amp or Pi proposes a plan, writes HTML or finishes code, it opens in your browser so you can annotate plans/specs, comment on local diffs or GitHub/GitLab PRs (Git, GitButler, Jujutsu, Perforce) and send the feedback straight back to the agent; built-in Ask AI and AI review comments, encrypted share links, a TUI and Herdr terminal plugin; one-line installer auto-configures hooks, skills and slash commands; GitHub |
| PentAGI | VXControl | Open-source (MIT), self-hosted autonomous AI penetration-testing platform: a team of specialist agents plans and runs tests in an isolated Docker sandbox with 20+ built-in tools (nmap, Metasploit, sqlmap and more), keeps long-term memory and an optional Neo4j knowledge graph, and produces vulnerability reports; web UI plus REST/GraphQL APIs, Docker Compose deployment and support for OpenAI, Anthropic, Gemini, Bedrock, Ollama and other OpenAI-compatible LLMs; GitHub |
| HeyDeer | ETOILEE PTE. LTD. (submitted by HeyDeer's maintainer) | HeyDeer is an AI code review service for GitHub pull requests that verifies findings before posting them and creates separate fix PRs from selected findings. Its recurring free plan includes private repositories, with 1,000 credits in the first month and 300 credits monthly thereafter. |
| farol | Winnin | Free, open-source (MIT) CLI (npx @winninjs/farol) and GitHub Action that runs after your AI code review and tells the human approver how risky a pull request is to merge: it builds a declaration graph with the real compilers (TypeScript type checker, go/types, Python ast) to list every page, API operation, RPC method, job and script the change reaches with call chains, then gives a deterministic verdict (fast track, quick review, careful review) and 0-100 risk score from reach, diff coverage, security findings, contract breaks (GraphQL, FastAPI, Go handlers, gRPC) and dependency changes; interactive HTML report, per-push incremental reports, fails closed as unknown, never executes or uploads your code and never blocks a merge; GitHub |
| no-mistakes | kunchenguid | Open-source (MIT) Go CLI and TUI that puts a local git proxy in front of your real remote: git push no-mistakes (or the /no-mistakes agent skill) spins up a disposable worktree and runs an AI-driven review → test → docs → lint → push → PR → CI pipeline using your own coding agent (Claude Code, Codex, Grok, Rovo Dev, OpenCode, Pi, Copilot, Antigravity, or Cursor/Devin via ACP, with ordered fallbacks); safe mechanical fixes are applied automatically, anything touching intent is escalated for you to approve, fix or skip, and the branch is only forwarded and a clean PR opened once every gate is green, with CI failures auto-repaired; macOS, Linux and Windows; GitHub |
🧪 AI Testing & Quality Assurance
Tools designed to autonomously generate, execute, and fix tests.
| Tool | Company | Notes |
|---|---|---|
| GitAuto | GitAuto | Automatically writes, runs, and fixes your unit tests, so you can keep shipping confidently |
| AgentDiff | AgentStatus | Catch the regression CI couldn't, before merge. |
| Agent QA | Vostride | The self-improving QA agent for software teams. It creates and runs natural-language web and mobile tests, retains test memory, and adapts flows when interfaces change. |
| e2e | TesterArmy | Open-source (Apache-2.0) agentic end-to-end testing framework for web, iOS and Android: TypeScript tests mix natural-language agent goals (agent.act, agent.assert) with deterministic locators and assertions, agent steps verified by a later assertion are recorded and replayed with no model calls until the app changes, web runs on Chromium/Firefox/WebKit via Playwright and mobile on simulators/emulators; BYO model, API key or ChatGPT/GitHub Copilot/SuperGrok subscription, an e2e mcp server lets Claude Code or Cursor drive a real browser or simulator, plus a GitHub PR-comment reporter; npx e2e init, pre-1.0.; GitHub |
| iFixAi | iFixAi | Open-source (Apache-2.0) independent auditing CLI for AI agents: ifixai setup then ifixai run executes 32 inspections across five pillars against any provider or your agent's real endpoint, graded by a self, independent or multi-judge ensemble into an A–F scorecard with JSON/Markdown reports for CI; also ships as a Claude Code/Codex plugin and a uvx ifixai install slash-command skill for Cursor, VS Code/Copilot, Windsurf, Cline, Continue, Gemini and Zed; GitHub |
| ARTEMIS | Open-source (Apache-2.0) Python agent that turns natural-language instructions into Android automation on real devices or emulators for end-to-end testing and exploratory/stability runs: a reactive observe-and-act loop (typically 3–5 s per step) with element-index, coordinate and visual targeting, Logcat and screenshot capture, a local web console with live screen mirroring and execution replays, and a built-in MCP server (artemis mcp --install all) plus testing rules so Antigravity, Claude Code, Codex, Cursor and Windsurf can drive the phone; reports 99%+ task completion on Google Research's AndroidWorld benchmark; one-click ./start.sh installs ADB, scrcpy, FFmpeg and uv dependencies |
|
| Promptfoo | Promptfoo (part of OpenAI) | Open-source (MIT) CLI and library for evaluating and red-teaming LLM apps and agents (npx promptfoo@latest, brew install promptfoo or pip install promptfoo): declarative test cases and assertions compare prompts and models side by side (OpenAI, Anthropic, Azure, Bedrock, Ollama and more), red teaming and vulnerability scanning probe agents and RAG pipelines, results run in CI/CD and a local web viewer, and code scanning reviews pull requests for LLM-related security issues; Promptfoo joined OpenAI in March 2026 and remains MIT-licensed; GitHub |
🧰 General-Purpose AI Assistants (with Strong Coding Capability)
Not IDEs, but widely used for coding tasks via chat.
| Tool | Company | Notes |
|---|---|---|
| Claude | Anthropic | Top-rated for complex debugging, architecture, and large codebase reasoning |
| ChatGPT | OpenAI | Versatile everyday coding; GPT-5 in 2026; broad language support |
| Gemini | 1M token context; multimodal; strong for Google ecosystem teams | |
| DeepSeek | DeepSeek | Competitive open-weights models (V3/R1); excellent coding benchmarks |
| Perplexity | Perplexity | AI search engine; excellent for technical research and API discovery |
| nenva | Anjadhe | Source-available (PolyForm Noncommercial) privacy-first personal AI assistant for macOS: email, calendar, tasks, notes, journal and investments live in its own apps and the AI comes to the data; new installs use hosted nenva cloud (zero data retention, free monthly allowance), or switch to a local open-weight model on 8 GB+ Macs, your own server or your own OpenAI/Anthropic key; free, no account; extensible via coding agents pointed at ~/nenva/apps/; GitHub |
| Odysseus | odysseus-dev | Open-source (AGPL-3.0) self-hosted AI workspace (Docker Compose) for chat and agents with local or API models, tools, MCP, files, shell, skills and memory, plus deep research, blind side-by-side model comparison, a hardware-aware model "Cookbook", an AI document editor, email triage, notes, tasks and CalDAV calendar with scheduled agent tasks; GitHub |
| Hob | Frigade | Open-source (MIT) desktop AI assistant for Apple silicon Macs, Windows and Linux that runs open-weight models on your own computer through llama.cpp: it checks your hardware and downloads a model that fits (Gemma 4 E4B or 26B, or gpt-oss 120B on workstations), then works on attached files and project folders, runs sandboxed Python, reads PDFs, scans and screenshots, draws charts and searches the web through a metasearch engine running locally; asks before changing anything in your folders, offers Undo for every task, needs no account and sends no telemetry, and Web access can be switched off to stay fully offline; extensible with Agent Skills, Python tools and MCP connectors; GitHub |
🔍 AI Codebase Knowledge & Generation
Tools for understanding large repositories or generating SDKs/APIs.
| Tool | Company | Notes |
|---|---|---|
| Greptile | Greptile | AI that ingests massive codebases and answers deep architectural questions via chat or API |
| Fern | Fern | Generates SDKs and developer documentation natively from your API definitions |
| Remio | Remio | Local-first AI memory and personal knowledge base desktop app. Remio parses files, webpages, recordings, emails, messages, images, and notes into local indexes and vectors so coding agents can retrieve focused personal context through the Remio CLI/skill instead of repeatedly scanning folders or loading whole documents into prompts. The CLI/skill require the Remio desktop client. |
| ContextStream | ContextStream | Shared project memory for Cursor, Claude Code, Codex, Grok, and the rest: one query, same memory across tools. Hosted MCP + npm package with semantic code search, knowledge graph, and decision tracking.; GitHub |
| OKF Agent Memory | okf-memory | Git-native persistent memory for AI coding agents that implements Google OKF v0.2. It stores project knowledge as plain Markdown with YAML frontmatter inside the repo, provides sub-300µs in-memory BM25 search, and exposes a built-in MCP server, all with zero external dependencies.; GitHub |
| Tseha | Tseha | Hosted MCP server that serves a team's design system and coding standards (components, design tokens, conventions) to Claude Code, Cursor and other MCP clients, so agents read the team's authored source instead of guessing from training data. Standards are curated and versioned by admins rather than accumulated at runtime; Figma token sync, npm/ZIP package import and semantic component search on paid tiers.; GitHub |
| Brain Scanner | Kevin Lozada Santos (submitted by his authorized AI assistant) | Brain Scanner gives you and your coding agent a shared project map to follow callers and dependencies before changing code, with a source-linked Codex example and first-map steps at https://github.com/kevin-lozada-santos/brain-scanner-walkthrough . It is a hosted MCP service with OAuth and a private implementation, currently free in open beta; maps can miss relationships and client compatibility varies. |
| codebase-memory-mcp | DeusData | Open-source (MIT) single-binary code intelligence MCP server that indexes a repo into a persistent knowledge graph with tree-sitter (162 languages, hybrid LSP type resolution for major languages) and exposes 17 tools (search, call tracing, architecture, impact analysis, dead code, Cypher queries) so agents answer structural questions in a few thousand tokens instead of grep/read loops. install auto-configures Claude Code, Codex, Cursor, OpenCode and 40+ other clients; built-in 3D graph UI.; GitHub |
| Gortex | zzet | Open-source (Apache-2.0) local code-intelligence engine for coding agents: indexes code into a persistent graph, connects multiple repositories and their API contracts, and exposes symbol search, call chains, impact analysis and graph-grounded review via MCP, CLI and web UI; single binary for macOS, Linux and Windows. |
| GitDiagram | Ahmed Khaleel | Open-source (MIT) web app that turns any public or private GitHub repository into an interactive AI-generated architecture diagram with a streamed explanation: click any component to jump to the linked file or directory, export PNG or copy the Mermaid source; replace hub with diagram in any GitHub URL; newer explainer mode renders a ~1-minute narrated video of the codebase (MP4, landscape or 9:16); Next.js + Bun, self-hostable with your own OpenAI/OpenRouter key; GitHub |
| jevgrep | David Zhang (dzhng) | Open-source (MIT) research CLI for coding agents: ask a natural-language question about a repository and jg uses TypeSafe's Jev model to judge relevance across folders, files and declarations, returning relevant files, reading leads and verbatim source excerpts in one stdout response; ships an agent skill (jg skill) so Claude Code, Codex and other agents use it for context gathering; BYOK via Vercel AI Gateway, TypeSafe, OpenRouter or OpenCode Zen |
| CodeGraph | Colby McHenry | Open-source (MIT), 100% local code-intelligence engine with a Rust kernel that pre-indexes a repo into a knowledge graph of symbols, call edges and dependencies (30+ languages, framework-aware routes, iOS/React Native bridging) and auto-syncs on every file change; agents get one codegraph_explore MCP tool that returns the relevant source, call paths (including dynamic-dispatch hops grep can't follow) and blast radius in a single call, plus a codegraph affected CLI for change impact; codegraph install wires it into Claude Code, Cursor, Codex, OpenCode, Hermes Agent, Gemini CLI, Antigravity, Kiro and GitHub Copilot; npx @colbymchenry/codegraph; GitHub |
| GitNexus | Abhigyan Patwari | Source-available (PolyForm Noncommercial) code-intelligence engine that indexes a codebase into a knowledge graph (dependencies, call chains, functional clusters and execution flows) and exposes it to Cursor, Claude Code, Codex, Windsurf, OpenCode and other agents through 19 MCP tools (hybrid search, 360° symbol context, blast-radius impact, git-diff change detection, coordinated rename, Cypher, cross-repo groups, opt-in PDG/taint queries); npx gitnexus analyze also installs agent skills, Claude Code hooks and AGENTS.md/CLAUDE.md, with a zero-server browser Web UI for graph exploration and repo chat plus LLM wiki generation; GitHub |
| code-review-graph | Tirth Kanani | Open-source (MIT) local code knowledge graph for AI code review: Tree-sitter builds a persistent SQLite map of functions, classes, imports, calls and inheritance, re-parses only changed files and their dependants, and serves compact review context over 30 MCP tools so assistants read only what a change touches (benchmarked ~65x median per-question token reduction); code-review-graph install configures Claude Code, Codex, Cursor, Gemini CLI, Antigravity, Windsurf, Zed, Kiro, Copilot, Hermes and more, plus a GitHub Action that posts risk-scored PR comments with an optional merge gate, interactive graph visualisation and wiki export; pip install code-review-graph; GitHub |
| Repowise | Repowise | Open-source (AGPL-3.0, commercial license available) local codebase index for coding agents: repowise init builds a call graph across 26 languages plus git history, tests, docs, ownership and architectural decisions without an API key, and repowise serve exposes a dashboard and MCP server so Claude Code, Codex, Cursor and other agents can ask what calls a symbol, what breaks if it changes, which tests a diff exercises, what code is dead and why it was built that way; auto-generated module docs update incrementally on each commit; install with uv tool install repowise or pip install repowise; GitHub |
| XERJ | XERJ | Open-source (Apache-2.0) local Rust search engine for coding agents: xerj autoindex <folder> indexes code, docs, logs, PDFs, SQLite and CSVs into typed queryable indices so agents query ranked file:line passages (xerj search, xerj def go-to-definition) instead of grepping whole files; its main use is "reference coding": the agent clones and indexes the open-source repos closest to the task and reuses how they solved it (project case study: 2.7x fewer output tokens than a grep-driven agent at the same solve rate); xerj init wires it into Claude Code and Cursor as an MCP server plus skill, and xerj gain reports measured savings; install with the one-line script at https://xerj.org/get (macOS/Linux); GitHub |
| Graphify | Graphify Labs | Open-source (Apache-2.0) /graphify skill and CLI (uv tool install graphifyy, then graphify install) that maps a codebase plus its docs, SQL schemas, configs, PDFs, images and video into a queryable knowledge graph for Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot and 20+ other assistants: code is parsed locally with tree-sitter (deterministic, no LLM, no vector store), every edge is tagged EXTRACTED or INFERRED, and each run writes an interactive graph.html, a GRAPH_REPORT.md and a reusable graph.json that agents query instead of grepping; also ships an MCP server; optional paid Graphify Cloud (14-day trial) keeps the graph live across repos with code review and Sentry/Jira connectors; GitHub |
| Graphicle | Graphicle AI | Open-source (Apache-2.0) single Rust program that gives coding agents a code graph plus project memory (brew install graphicle-ai/tap/graphicle, then graphicle setup): indexes the repo into a code graph, remembers docs, decisions and past agent sessions, and answers with cited file-and-line excerpts over MCP for Claude Code, Cursor, Codex, VS Code and Claude Desktop; graphicle ask from the terminal, graphicle open dashboard, graphicle graph affected impact checks, Git hooks keep the index current, and graphicle disconnect undoes setup. Runs locally with no API key, account or telemetry; positioned as a Rust port of Graphify with memory |
🛠️ Developer Productivity & Workflow
AI-powered tools for managing context, snippets, and developer documentation.
| Tool | Company | Notes |
|---|---|---|
| Pieces for Developers | Pieces | AI-powered context management; "Artificial Memory" for your workflow; snippet management |
| agenttrace | luoyuctl | Open-source observability and audit trail for AI coding agent sessions |
| Vibe Kanban | Vibe | Project management and productivity platform integrated with AI |
| AI-MEMORY | AkitaOnRails | Open-source (MIT) long-term memory for coding agents: quit Claude Code mid-task, resume in Codex, Gemini CLI, OpenCode, or Cursor in the same directory without re-explaining architecture, failed approaches, or open questions; cross-agent handoffs via MCP and hooks |
| smolvm | smol machines | Open-source (Apache-2.0) embeddable, portable, branchable microVM for safely running coding agents and untrusted code locally: boots OCI images in a hardware-isolated VM in under 200ms, with snapshot/branch, sandboxed shell, and Claude Code / Codex / Pi integrations |
| coop | Trail of Bits | Open-source (Apache-2.0) Rust CLI from Trail of Bits that runs Claude Code and Codex inside disposable, isolated VMs (Firecracker microVMs on Linux/KVM, Lima + Apple Virtualization.framework on macOS) so agents get full tool access (Docker, git, compilers, bypassPermissions by default) with the VM as the blast radius; coop up / coop claude / coop codex sync the project in, forward API keys (or keep them host-side via proxy mode), inject CLAUDE.md, MCP config and devcontainer.json; documented trust model and multi-instance support |
| SPACE | Perplexity | Open-source (Apache-2.0) Rust sandbox platform for agentic code execution, open-sourced by Perplexity as the sandbox layer behind Perplexity Computer: an HTTP API creates Linux sandboxes from templates as isolated Firecracker microVMs with their own writable filesystem, runs command sessions and streams output, and can keep, pause, suspend and wake sandboxes; Kubernetes-based deployment with warm pools, base snapshots, operator egress profiles, an operator CLI (space-admin-cli) and an E2B Compatibility Wrapper so existing E2B Python clients can use SPACE as their backend; local setup runs in a Kind cluster on a Linux x86_64 host with KVM; Docs |
| LayerFS | Ephemeral AI Lab | Open-source (MIT) Rust filesystem for AI coding agents: each agent mounts a complete, writable Workspace (FUSE view plus its own overlay) over shared committed history without copying the base, commits what it changed as deduplicated checkpoints, and can branch, rewind or merge, with conflicting Commits refused unless a .lww rule applies; one SQLite-file Store shared by daemons and the JavaScript SDK, no database server; 1.0.0 published benchmarks (fs-bench, pjdfstest, fsx, fsstress) and runs today through the dsh-layerfs plugin for DeepSeek Harness; Store has no crash or power-loss durability yet; GitHub |
| Claude Code Templates | Daniel Avila | Open-source (MIT) CLI (npx claude-code-templates) and web catalog for configuring Claude Code: installs agents, commands, hooks, MCP integrations, skills, and settings, plus real-time session analytics, health checks, and conversation monitoring |
| wshobson Agents | Seth Hobson | Open-source (MIT) multi-harness plugin marketplace of agents, skills and commands for Claude Code, Codex CLI, Cursor, OpenCode, Antigravity CLI, GitHub Copilot and Pi: 94 plugins (202 agents, 185 skills, 105 commands) covering language stacks, code review, testing, infrastructure and security; install via /plugin marketplace add wshobson/agents, codex plugin marketplace add, or single skills with npx skills add / gh skill |
| Director | Director | Native macOS workspace that places coding agents on a canvas for parallel planning, building, reviewing, and isolated Git worktrees |
| Claudete | Claudete | macOS control panel for running Claude Code, Codex, and Grok agents side by side with grids, shells, usage tracking, and broadcast prompts |
| PandaCode | AICodeKing | Free, open-source (MIT with a fork-attribution clause) native desktop workspace for coding agents built with Tauri 2 and Rust (macOS, Windows, Linux): runs Claude Code, Codex, OpenCode, Kimi Code, GitHub Copilot CLI, Antigravity, Grok Build, Cline, Pi and more side by side with projects, Git and per-task worktrees, a terminal, an embedded browser, a task board, architect-plus-workers agent teams, cloud agents on your own Daytona/E2B/Modal/AWS/Cloudflare account, and bots on Telegram, Slack or Discord; no account or telemetry |
| Fractal OS | Nubler Digital | Workspace for humans and agents with persistent memory, autonomous skills, connected tools, and adaptive interfaces across macOS, Linux, Windows, and web |
| Zuse | Zuse | Coding-agent workspace for running agents side by side, carrying context between them, and isolating tasks in Git worktrees |
| Cydonia | Crabtalk | Desktop workspace for ACP coding agents that turns sessions into durable on-disk artifacts such as articles, boards, and tables |
| AI Product Adoption Deck | aiproduct.cards | The AI Product Adoption Deck is a 124-page playbook of 104 cards that helps product teams diagnose and fix the specific moments where AI products lose users: empty prompts, trust gaps, broken correction loops, agents nobody trusts. It's built for PMs, designers, founders, and AI engineers shipping copilots, agents, and embedded AI features who need concrete patterns, not generic UX advice. Each card maps a symptom to a diagnosis, an action, and a workshop you can run with your team this week.; GitHub |
| CoderPlan | CoderPlan | LLM API Gateway with OpenAI-compatible interface. Pay-per-use access to Claude, GPT, Gemini, DeepSeek, Grok, and 100+ models. One-line config for Claude Code, Codex CLI, and Gemini CLI: switch providers without changing tools. |
| AIFlowLearn | AIFlowLearn / 木子霖 | AIFlowLearn is an AI learning-and-practice platform for AI engineering learners and builders. It turns AI frameworks, PDFs, articles, course materials, and open-source projects into structured learning collections, memory cards, and hands-on practice tasks so developers can move from reading to reviewable, repeatable engineering practice. |
| Faryo | Snailflyer | Faryo is a lightweight phone/desktop workbench for the same tmux-backed Codex CLI, Claude Code, or shell session. It keeps the terminal process as the source of truth while the phone/browser handles compact output, short input, approve/interrupt, and handoff. Same-session proof for the current release: https://github.com/Snailflyer/faryo/releases/tag/v1.0.7; GitHub |
| Tree Ring Memory | TerminallyLazy | Open-source local-first memory lifecycle layer for AI coding agents. The Rust CLI/TUI stores project memories in SQLite/FTS and supports recall, forgetting, audit, consolidation, evidence capture, and DOX/Revolve adapters so agents retain decisions and lessons without transcript dumps.; GitHub |
| OpenAgentRelay | ShakespeareLabs | Open-source CLI for sharing a restricted local coding-agent command with teammates or other agents over a trusted LAN, with keyed access, target verification, structured JSON output, and scriptable exit codes. Source code, prompts, dependencies, and credentials stay on the publisher machine.; GitHub |
| Better Agent | Ofek Ron | Local web workspace for launching and supervising native Claude, Codex, and Gemini coding-agent sessions with parallel delegation, persistent state, approval gates, file access, and restart recovery. It is source-available and free for non-commercial use; commercial use requires separate permission. Disclosure: I maintain Better Agent. “Freemium” is the form’s closest available pricing option. AI-assistance disclosure: this submission was drafted by Codex under the maintainer’s authorization and reviewed in Better Agent. |
| Tura | Tura-AI | Tura is a local, open-source execution layer for coding agents that groups repository inspection, edits, builds, tests, linting, and media inspection into fewer model turns. It publishes reproducible DeepSWE and full-repository rewrite benchmarks comparing its macro execution modes with Codex CLI Medium and High. Disclosure: I maintain Tura.; GitHub |
| fractal | Plasma AI | Open-source hierarchical coding-agent orchestrator for Claude Code, Codex, Grok Build, OpenCode, and Oh My Pi, with recursive delegation, per-node Git worktrees, configurable limits, and a live terminal UI for monitoring and steering.; GitHub |
| WolfMarkDown | WolfMark | WolfMarkDown is an AI agent skill for turning messy, AI-generated research, conversation or Markdown into clean, production-ready documents using semantic judgement plus deterministic verification. It repairs structure, preserves technical details, and validates the final file with Prettier, markdownlint, GFM parsing, integrity checks, and idempotence.; GitHub |
| Codex Quota Overlay | cpys (maintainer; independent community project) | An independent Windows and macOS overlay that shows remaining Codex quota, next reset time, and available reset credits beside the active Codex conversation. It reads the documented local rate-limits method, stays out of the way, and does not capture screenshots, conversations, or browser cookies.; GitHub |
| DeskCue | Alexander Kornev | DeskCue is an open-source, local-first workspace for reviewing and steering coding-agent work. It keeps sessions, git diff, files, live preview, notifications, and the next prompt accessible from another browser or phone while the agents and workspace stay on your machine.; GitHub |
| SandBase CLI | SandBase | Open-source CLI and MCP server that gives developers and AI coding agents one interface to discover and call 2,000+ AI models and APIs. It supports structured JSON output, 25 AI clients, local/remote MCP transports, and six MCP tools.; GitHub |
| AIPM | Abhishek Srivastava | Open-source registry and CLI for publishing, versioning, discovering, and installing reusable AI skills across Codex, Claude Code, and Cursor.; GitHub |
| Superagent | pungme | Open-source (MIT) macOS desktop app that gives Claude Code and Codex a real browser to drive, an iOS Simulator to install and screenshot apps in, and a phone companion app for remote monitoring.; GitHub |
| XiuRouter | XiuLab Inc | XiuRouter is a usage-based model API gateway for coding agents and developer tools. It supports OpenAI Chat Completions and Responses, Anthropic Messages, and Gemini GenerateContent, with scoped API keys, public pricing, and request-level usage records. |
| QwenCloud | Alibaba Cloud | QwenCloud provides model APIs and a Credits-based Token Plan subscription for coding and agent workflows. It supports OpenAI- and Anthropic-compatible APIs for tools such as Claude Code, Cursor, Qwen Code, and Codex. |
| OrcaReplay | Continuum AI | A record/replay/fork debugger for AI coding agents rather than an agent itself: it records a run of Claude Code, Codex, opencode, Qwen Code or Cursor below the harness, so model traffic, shell exit codes, per-turn file changes and MCP calls land on one timeline. The recording replays offline byte-for-byte with the network off, or forks from any checkpoint onto a different model to compare what each would have decided from the same state.; GitHub |
| agentplay | Entelligence AI | Open-source (Apache-2.0) local session replayer for coding agents (uvx agentplay): finds Claude Code, Codex, Cursor, GitHub Copilot (CLI and VS Code), OpenCode and Kiro transcripts already on your machine and turns each into a redacted radial map and playable timeline of files read and edited, failed commands and retries, MCP servers, subagents, context fill and the PRs or commits produced; no account or API key and session data stays local; GitHub |
| Scrawl | Kit Langton | Open-source (MIT) screen-capture tool for showing a coding agent a problem instead of describing it: press ⌥⇧S, talk, click and draw, and Scrawl packs the screen, voice and marks into an agent-readable bundle (annotated contact sheet, keyframes, time-aligned transcript and scrawl.json manifest); scrawl CLI plus an agent skill (bunx skills add kitlangton/scrawl) lets Claude Code, OpenCode or Codex zoom into moments, and captures can be sent straight into an OpenCode session; macOS (Apple silicon) via Homebrew, with Windows and Linux previews and cabled-iPhone recording; GitHub |
| Magic Patterns | Magic Patterns | AI product-design tool for generating high-fidelity UI, importing design systems, collaborating, and exporting production-ready code |
| Builder.io Visual Copilot | Builder.io | Design-to-code and AI development workflow grounded in existing frameworks, design tokens, components, and backend integrations |
| Jackalope | Jackalope Digital LLC | A desktop workspace for coding agents that brings projects, agent tasks, isolated Git worktrees, project context, recurring tasks, account profiles, usage reporting, and code review into one place. It is currently in early access with a waitlist. |
| Cabinet | Cabinet AI | Free, self-hosted AI workspace that combines company knowledge, files, AI teams, live apps, dashboards, tasks, and team chat while letting teams bring their own model providers |
| PI-Desktop | vastsa | Local-first desktop workspace for AI coding agents: Electron + Rust host around the Pi agent harness with projects, models, tools, long-running sessions, and user-installable plugins |
| Ruflo | ruvnet | Open-source agent meta-harness for Claude Code and Codex: one npx ruflo init adds 100+ specialized agents, coordinated swarms, self-learning memory, and federated communication across machines |
| Context Mode | mksglu | MCP server that keeps raw tool output out of the context window: sandboxed execution (~98% context reduction), SQLite/FTS5 session memory with BM25 recall, and routing enforcement across 17 agent platforms |
| Worktrunk | Worktrunk | Open-source Rust CLI (wt) for Git worktree management built for running Claude Code, Codex, and other agents in parallel: branch-addressed worktrees, create-and-launch-agent in one command, hooks, LLM commit messages, and a one-shot squash/rebase/merge/cleanup workflow; GitHub |
| RTK | RTK-AI | Open-source Rust CLI proxy ("Rust Token Killer") that filters and compresses shell command output before it reaches a coding agent's context; installs a Bash-rewrite hook for Claude Code, Codex, Gemini CLI, Cursor, Cline, Pi, Hermes, and more; GitHub |
| Gentle-AI | Gentleman Programming | Open-source (MIT) configurator that layers persistent memory, spec-driven development, curated skills, MCP servers, personas, and bounded review onto Pi, OpenCode, Claude Code, Codex, Cursor, Copilot, Gemini CLI, and Kilo Code without agent lock-in; GitHub |
| ECC | Affaan Mustafa | Open-source (MIT) agent harness system that installs a plan → test → implement → review → verify → remember loop into Claude Code (native ecc@ecc plugin), Codex, and Kimi Code, with adapters for Cursor, OpenCode, Gemini, Zed, and Copilot: 68 agents, 292 skills, hooks, memory, continuous learning, and AgentShield security scanning; GitHub |
| Atlas | pacifio | Open-source (MIT) macOS desktop app billed as "source control for coding agents": runs Claude Code, Codex, its own agent, and any ACP-registry agent side by side, links every commit back to the session, prompts, and tool calls that produced it (checkpoints), and shares on-device semantic memory across agents; GitHub |
| Archify | tt-a1i | Open-source (MIT) agent skill for Cursor, Claude Code, Codex CLI, and OpenCode that turns a codebase or system description into validated, interactive architecture, sequence, data-flow, and lifecycle diagrams: agents emit typed JSON IR, Archify deterministically compiles it into self-contained HTML/SVG with Before/Delta/After comparisons for reviewing architecture changes; GitHub |
| workmux | raine | Open-source (MIT) Rust CLI that pairs Git worktrees with tmux (or kitty, WezTerm, Zellij) windows so multiple coding agents can run in parallel: one-command add/merge lifecycle, agent-status dashboard and sidebar, /worktree delegation skill, post-create hooks, container/VM sandboxing, and LLM-generated branch names; GitHub |
| Spec Kit | GitHub | Open-source (MIT) toolkit for Spec-Driven Development with any AI coding agent: specify CLI (uv tool install specify-cli) scaffolds a project, then /speckit-constitution → specify → plan → tasks → implement → converge skills carry a spec through implementation; bundled bug-fixing and idea-assessment processes; 30+ agent integrations (Copilot, Claude Code, Codex, Gemini CLI, Cursor); GitHub |
| Agent Skills | Addy Osmani | Open-source (MIT) pack of 25 production-grade engineering skills for coding agents: /spec, /plan, /build, /test, /constraints, /review, /webperf, /code-simplify, /ship slash commands mapped to the SDLC, with auto-activating skills (TDD, API design, frontend UI, code review) and anti-rationalization gates; installs into 70+ agents via npx skills add addyosmani/agent-skills; GitHub |
| 9Router | decolua | Open-source (MIT) local AI gateway (npm install -g 9router) that exposes one OpenAI-compatible endpoint for Claude Code, Codex, Cursor, Cline, Copilot, Antigravity, OpenCode, and OpenClaw: 3-tier auto-fallback (subscription → cheap → free) across 40+ providers, quota tracking, multi-account round-robin, format translation, and built-in RTK/Caveman token compression; GitHub |
| i-have-adhd | ayghri | Open-source (MIT) agent skill/plugin that stops coding assistants from burying the answer: 10 rules (lead with the next action, number steps, restate state, cap lists at 5, no preamble/recap/closers); installs into Claude Code, Codex, Gemini CLI, Copilot, Antigravity, Hermes, and OpenCode via plugin marketplace or npx skills add; GitHub |
| Oh My Hermes | rlaope | Open-source (MIT) operating layer for Hermes Agent (omh setup, via Homebrew/Bun/npm or install script): 116 installable skills grouped into seven capability families, nine ulw-* workflows for planning, research, creation, operations, and project memory, per-category model chains, and a Maestro lane that hands coding work off to Claude Code or Codex with explicit evidence boundaries; works in the Hermes CLI, Desktop, and messenger surfaces; GitHub |
| Superpowers | obra | Agentic skills framework and software development methodology that installs into Claude Code, Codex, Cursor, Antigravity, and other agents as composable skills |
| BlaBla | Petro Yakovyshyn (independent; I maintain BlaBla) | Executable project memory for coding agents: an MIT-licensed Rust CLI that keeps project intent as contracts an agent queries instead of re-reading chat history. behavior.bla declares observable state, actions, postconditions and invariants, checked by a coverage-guided campaign that shrinks a failure to a minimized counterexample; structure.bla declares modules, symbols, dependencies and literal collection membership, evaluated statically by a Python ast provider in an isolated interpreter that never imports project code. blabla status and blabla explain <rule> are the read path, and blabla finish is a completion gate that exits 0 only when every active layer is GREEN. Disclosure: I maintain BlaBla and am submitting it myself. BlaBla runs no model of its own: it is deterministic infrastructure built for AI coding agents, like the worktree, context and memory CLIs already in this category. It is v0.5.0-alpha, with a Python-only structure provider in this release; close this if that is too early for the directory.; GitHub |
| OpenSpec | Fission-AI | Open-source (MIT) lightweight Spec-Driven Development framework for AI coding assistants (npm install -g @fission-ai/openspec, openspec init): /opsx:explore → propose → apply → archive slash commands create a per-change folder with proposal.md, specs/, design.md, and tasks.md that you approve before code is written; plain-Markdown SHALL/WHEN/THEN specs live in the repo, Stores (beta) share specs across repos and teams; works with 20+ assistants (Claude Code, Cursor, Copilot, Windsurf, Cline, Codex, Amazon Q); GitHub |
| BrowserSkill | Tencent | Open-source (MIT) bsk CLI/daemon + Chrome/Edge extension that lets any shell-capable coding agent (Cursor, Claude Code, Codex, OpenClaw, CodeBuddy, Pi, Hermes Agent) drive your real, already-logged-in browser in a separate visible Agent Window: reuse real login state without test accounts, explicit tab borrow/return, built-in human-in-the-loop handoff for captchas and confirmations, full-page screenshots, bsk install-skill installs the harness skill; macOS, Linux, Windows; GitHub |
| Multica | Multica | Open-source, self-hostable (Docker Compose/Helm) workspace where AI coding agents show up on the issue board like teammates: assign an issue to Claude Code, Codex, Cursor, Copilot, Kimi, OpenCode or 20+ other agent CLIs, it runs on a daemon on your own machine, comments progress, raises blockers and hands the diff back for review; squads, skills, cron "autopilots", any Git host.; GitHub |
| Engram | Gentleman Programming | Persistent, agent-agnostic memory for coding agents: a single Go binary (SQLite + FTS5, no Node/Python/Docker) exposing 15 MCP tools (mem_save, mem_search, mem_context...), an HTTP API, CLI and TUI; one-command engram setup for Claude Code, Codex, Gemini CLI, OpenCode, Cursor, Windsurf, VS Code Copilot, Antigravity and more; optional Engram Cloud for teams.; GitHub |
| WebCodex | yyjeqhc | Self-hosted Server + Runner that lets ChatGPT, Claude and other MCP clients work on repositories that stay on your own machines: read/search files, guarded edits, Git, commands, tests and long-running jobs, with a Console to accept/reject work; webcodex share opens a Cloudflare Quick Tunnel, Desktop app pairs with the OpenAI Secure MCP Tunnel.; GitHub |
| shadcn lint | shadcn | Open-source (MIT) "agent-first" linter (@shadcn/lint) for Tailwind v4 design systems, as an ESLint 9 or Oxlint plugin: you declare what components may be restyled and rules like no-restyle, no-raw-colors, no-arbitrary-values, no-inline-styles, and no-unknown-classes emit errors that tell the coding agent the allowed fix (sizes, variants, theme tokens) read from your own components; shadcn/ui not required |
| Codenotch | vinzdg | Open-source (MIT) macOS app (Rust/Tauri Windows port in the same repo) that pins a small notch to a screen edge with one ring per coding-assistant usage limit (Claude Code, Cursor, Codex, Antigravity, GitHub Copilot, GLM, Grok and more); hover for limit windows and reset times; the ring doubles as a status light showing whether the agent is working, done or waiting on you. Reads the sessions already on your machine (Claude Code /usage, Cursor SQLite state, Codex rollout logs), signed DMG with Sparkle auto-update.; GitHub |
| Agent-Native | Builder.io | Open-source (MIT) TypeScript framework for building "agent-native" apps: define each capability once as an action and it is exposed to the UI, the agent (as a tool), HTTP, MCP, A2A and the CLI with the same validation and permissions. Ships agent chat, skills/memory, automations, agent teams, a Claude Code/Codex-style workspace, PostgreSQL/PGlite backend on any Nitro host; npx @agent-native/core@latest create my-app, plus open-source example agents (Clips, Plans, Design, Chat).; GitHub |
| Screenpipe | Screenpipe (Negentropy Labs, Inc.) | Search captured screen text and audio history, then give coding agents context from past debugging sessions, documentation, and meetings through MCP or a local API. Raw history stays local by default; configured cloud AI, transcription, sync, and integrations can send context off-device. |
| antiburn | antiburn | Free, local desktop app that checks coding agent sessions for common causes of token burn: sessions too deep, overpowered subagents, broken caching, and unused MCPs, skills, and tools. Supports Claude Code, Codex, Cursor, Copilot, Pi, and more.; GitHub |
| claude-mem | thedotmack | Open-source (Apache-2.0) persistent-memory plugin for coding agents: captures tool-usage observations during a session, compresses them into semantic summaries, and injects relevant context back into future sessions. Installs via npx claude-mem install or the Claude Code plugin marketplace, with adapters for OpenCode, Antigravity CLI, Cursor, Grok Bot and more; local observer is free, hosted CMEM Pro is optional.; GitHub |
| Oh My OpenAgent | Sisyphus Labs (code-yeongyu) | Source-available (SUL-1.0) multi-agent orchestration layer for OpenCode and Codex CLI: 11 specialised agents, 54+ lifecycle hooks, built-in websearch/context7/grep_app/LSP MCPs, Team Mode, /goal, and the ultrawork keyword that drives tasks to completion across Claude, GPT, Kimi and other models. Ships as an OpenCode plugin (Ultimate), a Codex plugin (Light), and a standalone omo CLI (beta); formerly oh-my-opencode.; GitHub |
| CLI-Anything | HKUDS | Open-source (Apache-2.0) framework and registry for making existing software agent-native: a 7-phase harness generator (as a Claude Code, Codex, Pi, OpenClaw, Cursor or Qoder skill) that wraps apps like Blender, GIMP, FreeCAD, Obsidian, n8n or Godot in structured-JSON CLIs with generated SKILL.md files, plus pip install cli-anything-hub to browse and install community-built CLIs from CLI-Hub.; GitHub |
| CC Switch | farion1231 | Open-source (MIT) cross-platform Tauri 2 desktop app that centrally manages Claude Code, Claude Desktop, Codex, Gemini CLI, Grok Build, OpenCode, OpenClaw and Hermes Agent configs: one-click switching between 50+ API-provider presets, unified MCP and Skills management with bidirectional sync, usage/balance tracking, endpoint speed tests, local routing with format conversion and failover, system-tray quick switch and cloud sync via Dropbox/OneDrive/iCloud/WebDAV; configs stored in SQLite with atomic writes; GitHub |
| magpie | yetone | Open-source (MIT) desktop, TUI and CLI model picker for Claude Code, Codex, Gemini CLI, OpenCode and other coding agents; edits each agent's native config and runs a local gateway translating OpenAI and Anthropic APIs so models from different providers appear across agents. Supports named profiles and macOS, Linux and Windows. |
| Hindsight | Vectorize | Open-source (MIT) agent memory system that makes agents learn rather than just recall: retain / recall / reflect API over PostgreSQL + pgvector with entity, temporal and semantic memory pathways, auto-consolidated observations and mental models; adds memory to any agent via a 2-line LLM-client wrapper, SDKs, MCP server or coding-agent integrations (Claude Code, Cursor), self-hosted with Docker or via Hindsight Cloud; state-of-the-art LongMemEval results; GitHub |
| Treg | Superdesign | "OpenRouter for agent tools": open-source (Apache-2.0 with additional terms), self-hostable tool registry/proxy that gives an agent one base URL and one token to call 2,800+ catalogued API endpoints across ~60 providers (SEO, social, enrichment, ads, scraping, image/video generation) priced per call with no provider signup, plus your team's own API keys, OAuth connections, vendor CLIs and SKILL.md bundles with credentials injected server-side so keys never reach the agent; GitHub |
| Impeccable | Paul Bakaus | Open-source (Apache-2.0) design skill for AI coding agents that started from Anthropic's frontend-design skill: npx impeccable install then /impeccable init writes PRODUCT.md, and 24 commands (craft, shape, critique, audit, polish, bolder, quieter, distill, animate, typeset, live, ...) give you a shared design vocabulary with the agent; 61 deterministic detector rules run via CLI or browser extension with no LLM, plus live in-browser variant iteration; installs into Claude Code, Codex, Cursor, Gemini CLI, Grok Build, Hermes and Copilot with provider-native edit hooks; GitHub |
| Chrome DevTools MCP | Google (Chrome DevTools) | Official Chrome team MCP server (npx -y chrome-devtools-mcp@latest) and CLI that lets coding agents control and inspect a live Chrome: record performance traces and extract DevTools insights, analyse network requests, read console messages with source-mapped stack traces, take screenshots and automate reliably via Puppeteer with automatic waiting; --slim and --headless modes, connect to a running Chrome, persistent profiles, Android debugging, and client guides for Antigravity, Claude Code, Cursor and VS Code; GitHub |
| Mobile MCP | Mobile Next | Open-source (Apache-2.0) Model Context Protocol server that lets coding agents drive native iOS and Android apps on simulators, emulators and real devices through one platform-agnostic API: accessibility-tree snapshots first (no vision tokens) with screenshot + coordinate fallback, taps/swipes/gestures, app install/launch, deep links, screen recording and structured data extraction; setup guides for Claude Code, Codex, Cursor, Cline, Copilot, Gemini CLI, Windsurf and more; npx @mobilenext/mobile-mcp; GitHub |
| OpenRig | Esoteric Labs | Open-source (Apache-2.0) multi-agent harness that runs Claude Code and Codex sessions together as one managed system: define an agent team (pods, edges, continuity policies) in YAML, boot it with rig up, and manage it through a local daemon, 40+ command CLI, MCP server and TUI/React dashboard built on tmux, with cross-agent messaging (rig send, rig chatroom), a task queue, snapshots and restore after reboot; npm i -g @openrig/cli; GitHub |
| Cua | Cua | Open-source (MIT) computer-use infrastructure for AI agents: Cua Driver operates apps on macOS, Windows and Linux via MCP or CLI without taking over the system cursor, Lume runs local macOS/Linux VMs on Apple Silicon, Cua Bench builds tasks and evaluates agents, and CUA-S1 provides small computer-use decision models; paid Cua Fleets provide isolated cloud desktops for agent workloads (YC-backed); GitHub |
| OpenShell | NVIDIA | Open-source (Apache-2.0) sandboxed runtime for autonomous AI agents: openshell sandbox create -- claude (or codex, opencode, copilot) runs the agent in an isolated sandbox where kernel-level controls confine files, syscalls and processes, every outbound connection passes a declarative YAML policy check, and credentials are injected only at approved endpoints so agents never see real keys; a privacy router keeps inference on local or self-hosted backends, and a formal policy prover flags risky access before changes are approved; Docker, Podman, microVM and Kubernetes compute drivers plus Python/TypeScript/Go/Rust SDKs; GitHub |
| agent-manager | YoanWai | Open-source (Apache-2.0) Go terminal UI for macOS, Linux and Windows via WSL2 that runs coding-agent CLIs side by side, each in its own persistent tmux session, launching your own installed Claude Code, Codex, OpenCode, Grok Build, Gemini CLI, Pi, Command Code or Hermes Agent unmodified so logins, subscriptions, config files and MCP servers carry over. It adds a live-status list of every session, optional per-session git worktrees, a full-file diff review whose line comments go back to the agent as one prompt, and a built-in MCP server that lets one agent spawn, message and wait on another session.; GitHub |
| Buzz | Block | Open-source (Apache-2.0), self-hostable workspace where humans and AI agents share channels, threads, DMs, canvases, voice huddles, Git hosting and YAML workflows; built on the Nostr protocol so every message, patch (NIP-34), review approval and workflow step is a signed event, and each agent gets its own cryptographic identity, channel memberships, encrypted memory and audit trail; agents run Claude Code, Codex, goose or any ACP-compatible harness via an agent-first buzz-cli; Rust relay with a Tauri desktop app for macOS, Linux and Windows and a one-click Railway relay deploy; GitHub |
| Beacon | Asymptote Labs | Open-source (MIT), local-first cross-harness memory layer for coding agents: captures full session traces (prompts, tool calls, commands, edits, approvals, MCP activity, tokens) from Claude Code, Cursor, Codex, OpenCode, Cline, Gemini CLI, Hermes and 20+ other local, CI and cloud agents into an OpenTelemetry-based event model, distils corrections, workflows and debugging patterns into reviewed memory, and serves it back to agents via MCP or Agent Skills; forwards to Splunk, Datadog, Elastic, Sentinel, S3 or GCS, with optional hosted Beacon Managed (search, analytics, SSO/RBAC); brew install beacon; GitHub |
| Hyperconsciousness | Louis Beaumont (maintainer) | MIT-licensed developer-alpha knowledge store that gives coding agents persistent context through a Rust CLI and MCP/HTTP server. It stores encrypted, signed, append-only records, synchronizes them across devices, and limits agent access through scoped, expiring grants.; GitHub |
| gstack | Garry Tan | Open-source (MIT) Claude Code skill pack from YC's Garry Tan that turns one agent into a virtual engineering team: chained slash-command roles (/office-hours, /plan-ceo-review, /plan-eng-review, /review, /investigate, /qa (drives a real browser), /cso security audit, /ship and /land-and-deploy), where each step writes artifacts the next one reads. Installs into ~/.claude/skills with a team mode that auto-updates for shared repos, and also targets Codex CLI, Cursor and OpenClaw (--host). |
| Beads | Steve Yegge | Open-source (MIT) bd CLI that gives coding agents a persistent, dependency-aware issue graph instead of markdown TODO plans: bd ready surfaces claimable work, bd update --claim / bd close release blockers, hash-based IDs avoid merge collisions across agents and branches, and old closed work is compacted to save context. Backed by Dolt (version-controlled SQL with branch/merge and remote sync); bd init / bd setup claude/codex/factory wire it into AGENTS.md and agent hooks; MCP server on PyPI.; GitHub |
| TUIOS | Gaurav Gosain | Open-source (MIT) Go terminal multiplexer and tiling window manager (Charm stack) that tracks coding agents: each agent pane reports working / needs_input / idle / done / errored via installed hooks (tuios integration install claude-code/codex/gemini-cli/opencode), one Inbox (ctrl+b i) collects approvals, questions and finished turns across sessions and machines, tuios fan starts one prompt in several agents in separate git worktrees, and tuios mcp plus a tmux shim expose the same surface to agents. Sessions survive restarts via daemon mode; Homebrew, AUR, Nix, Docker.; GitHub |
| OmniRoute | Diego Souza (diegosouzapw) | Open-source (MIT) self-hosted AI gateway (npm install -g omniroute, localhost:20128) that gives Claude Code, Codex, Cursor, Cline, OpenCode, Gemini CLI and other agents one OpenAI-compatible endpoint across 350+ providers (150+ with free tiers). "Combos" chain models with 19 routing strategies and an auto mode that fails over when quota runs out or a provider errors; omniroute run claude/codex/… launches CLIs pre-wired to the gateway; Docker, Electron, Termux and a VS Code Copilot Chat extension.; GitHub |
| Monty | Pydantic | Open-source (MIT) minimal, secure Python 3.14 sandbox written in Rust for running code written by AI agents ("code mode") without containers: a new sandbox starts in under 1ms from a worker pool, has no filesystem, environment or network, and reaches the host only through the functions and mounts you pass in; resource limits plus serializable snapshots that pause at a host call and resume later. Packages for Python (uv add pydantic-monty), JS/TS (@pydantic/monty) and Rust (cargo add monty); the commercial Full Monty runs the same workers behind a WebSocket with OS-level isolation; v1.0 stable in September 2026.; GitHub |
| gcx | Grafana Labs | Open-source (Apache-2.0) Grafana CLI built for you and your AI coding agent: structured JSON/YAML access to dashboards, alerts, SLOs, metrics, logs, traces, Synthetic Monitoring, k6, IRM and more on Grafana Cloud or Grafana OSS/Enterprise 12+, works with any agentic coding tool and ships a bundled Agent Skills pack (alert investigation, dashboard GitOps, SLO management, observability rollout), GitOps pull/push and Grafana Assistant integration on Cloud; brew install gcx; generally available since July 2026.; GitHub |
| agentgateway | Agentic AI Foundation (Linux Foundation) | Open-source (Apache-2.0) Rust proxy and gateway control plane for agentic traffic, hosted by the Agentic AI Foundation: one data plane for LLM, MCP, A2A, HTTP and gRPC with an OpenAI-compatible LLM gateway (budgets, spend controls, failover), MCP tool federation across stdio/HTTP/SSE/Streamable HTTP with OAuth, A2A agent-to-agent routing, guardrails, JWT/API-key/OAuth auth, CEL-based RBAC, rate limits and OpenTelemetry; runs standalone from YAML or on Kubernetes via Gateway API, with a built-in UI; GitHub |
| Weave Router | Weave | Open-source (Apache-2.0) model router for coding agents: a drop-in Anthropic/OpenAI/Gemini-compatible proxy that scores each agent request with an on-box embedder and sends it to the cheapest enabled model likely to succeed, with cache-aware switching, escalation back to a frontier model when a cheaper one stalls, and routing across your existing provider subscriptions; npx @weave-os/router wires Claude Code, Codex, opencode or pi to the hosted router, or self-host with BYOK.; GitHub |
| cf | Cloudflare | Open-source (Apache-2.0) "agentic CLI for the entire Cloudflare API": the command surface mirrors the full Cloudflare API (generated by Cloudflare's now open-sourced Forge SDK generator), with bespoke command search and steering so coding agents can find the command they need, JSON output by default (pretty-printed for humans, condensed for agents to save context), and a TypeScript cloudflare.config.ts configuration format with LSP support, starting with Workers and Vite as the default local dev server; open beta via npm i -g cf, near-daily releases.; GitHub |
| Taste Skill | Leon Lin | Open-source (MIT) anti-slop frontend skill pack for AI coding agents: npx skills add https://github.com/Leonxlnx/taste-skill installs portable skills (default design-taste-frontend v2 with VARIANCE/MOTION/DENSITY dials, a stricter GPT/Codex variant, redesign-audit, minimalist, brutalist, Google Stitch-compatible and full-output enforcement) plus image-generation skills for web/mobile comps and brand kits to hand to Codex, Cursor or Claude Code; GitHub |
| Skills Manager | Jay.TL | Open-source (MIT) desktop app that manages AI agent skills across 50+ coding tools (Claude Code, Codex, Cursor, Copilot, Gemini CLI and more) from one central library: install from Git repos, local folders, .zip/.skill archives or the skills.sh marketplace, group skills into presets, manage global and per-project workspaces with two-way sync, back up across devices, and let agents drive it to install or deploy skills instead of writing into each other's folders; GitHub |
| HyperFrames | HeyGen | Open-source (Apache-2.0) "write HTML, render video" framework built for coding agents: turns HTML, CSS, media and seekable animations (GSAP, Lottie, Three.js, CSS/WAAPI) into deterministic MP4s via the hyperframes CLI, with a Claude Code plugin and 21 agent skills (npx skills add heygen-com/hyperframes) for Codex, Cursor, Gemini CLI and others covering product-launch videos, PR-to-video changelogs, captions, motion graphics, slideshows and Remotion-to-HyperFrames migration; GitHub |
| BashCut | Nolan / dongnguyenvie | A native macOS video editor that coding agents can drive through CLI and MCP. Agent edits stay visible, atomic, and undoable in the timeline.; GitHub |
| Matt Pocock Skills | Matt Pocock | Open-source (MIT) set of small, composable agent skills for "real engineering, not vibe coding": /grill-me and /grill-with-docs alignment interviews before a change, /to-spec, /triage against your issue tracker and labels, /diagnosing-bugs, /code-review and /improve-codebase-architecture; install as a managed Claude Code plugin (claude plugin install mattpocock-skills) or as editable files for Codex and other agents via npx skills@latest add mattpocock/skills, then run /setup-matt-pocock-skills once per repo; GitHub |
| REA | morluto | Open-source (MIT) "Reverse Engineer Anything" MCP server and CLI that lets coding agents inspect apps without source code: native Mach-O/ELF/PE binaries via Hopper or Ghidra, JavaScript and Electron apps, .NET assemblies, Android APKs, firmware and websites, with evidence-backed explanations of how a feature works so the agent can build a version for your project; analysis runs locally, npx rea-agents setup registers it with Claude Code, Codex, Cursor, Gemini CLI, Windsurf, Devin, OpenCode, Copilot CLI and more; GitHub |
| Diagram Design | Cathryn Lavery (LittleMight) | Open-source (MIT) agent skill for editorial-quality diagrams in Claude Code, Codex, GitHub Copilot, Factory Droid, Pi and other Agent Skills hosts: 50+ diagram types (architecture, sequence, ER, state machine, dependency graph, UML class, database schema, Gantt, Sankey, Wardley and more) as self-contained HTML + SVG with no build step, brand matching by reading your website, optional accessible motion and redraws of Mermaid, draw.io or Excalidraw sources; npx skills add cathrynlavery/diagram-design or the Claude Code plugin marketplace; GitHub |
| DBX | DBX | Open-source (Apache-2.0) ~25 MB database client for 100+ databases (MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, ClickHouse, SQL Server, Oracle, vector stores and more) on desktop, Docker and web, with a built-in AI SQL assistant (generate, explain, optimize and fix SQL with safety checks; Claude, OpenAI, local or OpenAI-compatible models) and a separate Rust MCP server (npx @dbx-app/mcp-server) so Claude Code, Cursor, Windsurf and other agents can query databases through connections configured in DBX; GitHub |
| Obsidian Mind | Brenno Ferrari (breferrari) | Open-source (MIT) Obsidian vault template that gives coding agents persistent memory: hooks and slash commands (/om-standup, /om-dump, /om-incident-capture, wrap-up) for Claude Code with working Codex CLI and Gemini CLI support keep notes, decision records, people and project indexes linked across sessions, with optional QMD semantic search; install via shardmind install github:breferrari/obsidian-mind or clone the template; GitHub |
| GBrain | Garry Tan | Open-source (MIT) agent memory layer from YC's Garry Tan: stores explicit facts with sources, corrections and withdrawal, offers keyword then optional semantic retrieval, a cited synthesis layer and typed knowledge graph, and shares one brain across Claude Code, Codex, OpenClaw, Hermes and other harnesses via MCP (gbrain mcp expose over Tailscale) with installable skill packs; install with bun install -g github:garrytan/gbrain; GitHub |
| agent-browser | Vercel Labs | Open-source (Apache-2.0) fast native Rust browser automation CLI for AI agents: agent-browser open, snapshot (accessibility tree with @e refs), click, fill, get text, screenshot and recording commands let Claude Code, Codex, Cursor and other agents drive Chrome without Playwright or Node.js in the daemon; supports auth vaults, WebMCP tools, a chat mode and cloud browsers (Browserbase, Browser Use, Kernel, AgentCore); install via npm, Homebrew or Cargo, plus an agent skill (skills.sh/vercel-labs/agent-browser); GitHub |
| BMAD Method | BMad Code | Open-source (MIT) "Breakthrough Method for Agile AI-Driven Development": a set of agent skills and workflows (briefs, PRDs, architecture, epics/stories, bmad-build, review triage) that make decisions explicit and carry context forward from idea to working code, sizing the process to the change; installs via npx skills add bmad-code-org/BMAD-METHOD or the Claude Code and Codex plugin marketplaces and works with any skills-capable coding tool; GitHub |
| Tale | Ruler GmbH | Tale is an open-source project workspace where teams delegate tasks to coding agents such as Claude Code and Codex in persistent sandboxes, then review reports and deliverables together. Submitted by an AI assistant on behalf of Tale’s owner; the MIT Community edition can be self-hosted, with infrastructure and model costs separate.; GitHub |
| Screen Browser | MiCCA OÜ | Remote MCP server and companion skill that let Claude Code, Codex, Cursor or any MCP client record a narrated demo or tutorial video of a deployed web app: the agent writes a plain-language guide from the source code, validates it, starts the recording and returns the MP4.; GitHub |
| Headroom | Headroom Labs | Open-source (Apache-2.0) local context-compression layer for AI agents: compresses tool outputs, logs, JSON, code and RAG chunks before they reach the LLM, available as a Python/TypeScript library, a zero-code proxy (headroom proxy), an MCP server, and headroom wrap <agent> to route existing coding agents (Claude Code, Codex, Cursor, Aider, Cline and more) through it; originals stay cached locally for on-demand retrieval, plus cross-agent memory and headroom learn for writing fixes into CLAUDE.md/AGENTS.md; GitHub |
| CubeSandbox | Tencent Cloud | Open-source (Apache-2.0) self-hosted sandbox service for AI agents built on RustVMM and KVM: each sandbox is a hardware-isolated microVM with its own kernel, cold-starting in about 60 ms with under 5 MB overhead, E2B SDK-compatible (switch from E2B Cloud by changing one environment variable), with copy-on-write snapshots, clone and rollback, eBPF egress policies with a credential-injecting proxy, and single-node, Terraform or Kubernetes deployment; GitHub |
| Reqode | Almware Ltd. | Reqode helps software teams define whole features, give AI coding agents connected product and architecture context, and verify implementation against specifications. It connects requirements, data models, UI and API specifications, code, and tests, with AI assistants for analysis, architecture and test design and MCP access for external coding and QA agents. |
| MXC | Microsoft | Open-source (MIT) Microsoft eXecution Container: policy-driven sandbox for running untrusted code such as model output, agent tools and plugins on Windows, Linux and macOS behind one containment model: JSON container requests with filesystem, network-egress and UI/clipboard policies, backends from OS-native process sandboxes (ProcessContainer, Bubblewrap, Seatbelt, LXC, WSLC) to Windows Sandbox and microVMs, audit mode for authoring policies, and stable v1 SDKs for Rust (mxc-sdk), .NET (Microsoft.Mxc.Sdk) and Node (@microsoft/mxc-sdk) |
| Moli | Lexmount | Open-source (Apache-2.0 / MIT) Rust headless browser built for AI agents: runs real JavaScript, DOM, CSS, networking and storage but only computes layout and paint when an operation needs pixels, so moli fetch can return HTML, Markdown, JSON or a compact semantic tree cheaply, with screenshots and PDFs on demand; moli serve exposes CDP, WebDriver Classic and WebDriver BiDi on one endpoint so Playwright and Puppeteer connect directly, and bundled agent skills install it for coding agents; Linux, macOS and Windows; GitHub |
| Free Claude Code | Ali Khokhar (Alishahryar1) | Open-source (AGPL-3.0) local model proxy and launcher for coding agents (FCC; independent, not affiliated with Anthropic): one Admin UI connects 60+ free, paid, subscription (OAuth) and local providers (NVIDIA NIM by default) and fcc-claude, fcc-codex, fcc-pi, fcc-opencode, fcc-cline, fcc-aider and other launchers run 11 agents against them, with ordered fallback models across providers, per-tier (Opus/Sonnet/Haiku) routing, optional RTK and Headroom token reduction, proxy bearer-token auth, a browser UI for native Codex sessions, VS Code/JetBrains/Discord/Telegram clients and voice input via local Whisper; free-tier limits are set by each provider |
| Pairfob | arronKler | Free, Apache-2.0 phone companion for live Herdr coding-agent sessions: check progress, reply to Claude Code or Codex, handle confirmation prompts, and review git diffs from a mobile browser. The agents keep running on your macOS or Linux computer, which must stay awake and online; Herdr is required.; GitHub |
| Aident Loadout | Aident AI | Connect Codex, Claude Code, Cursor, ChatGPT and other MCP clients to 1,000+ apps and 400+ expert-built Skills through one reusable setup. Credentials stay in Aident Vault; every action is logged in Audit.; GitHub |
| agentmemory | rohitg00 | Open-source (Apache-2.0) persistent memory server for AI coding agents, installed with npx @agentmemory/agentmemory: an interactive setup wires Claude Code, Cursor, Codex CLI, Gemini CLI, GitHub Copilot CLI, OpenCode, Hermes, OpenClaw, pi and any MCP client, then captures observations into a knowledge graph with confidence scoring, decay/consolidation and hybrid BM25 + vector search; optional local embeddings, a real-time web viewer and REST/MCP APIs; publishes LongMemEval-S retrieval benchmarks (95.2% R@5) |
| TeamAI | Tencent | Open-source (MIT) CLI and agent skill (npm install -g teamai-cli) that keeps a team's AI assets in one Git repo (skills, rules, agents, hooks, MCP servers, env, models, learnings and a team wiki) and auto-syncs them into each teammate's Claude Code, Codex, Cursor, GitHub Copilot CLI, OpenCode, Kiro, Trae, Qoder, CodeBuddy and other agents; project- or user-scope installs, protected-branch publishing, usage/session insights (beta) and a dashboard |
| Agent Toolkit for AWS | AWS | Official open-source (Apache-2.0) AWS kit that gives coding agents the tools, knowledge and guardrails to build, deploy and manage apps on AWS: plugins for Claude Code (on the official claude-plugins-official marketplace), Codex and Cursor (aws-core, aws-agents, aws-data-analytics, aws-agents-for-devsecops and aws-startup-advisor) bundle the managed AWS MCP Server (300+ services through one endpoint, sandboxed Python script execution, live docs search, IAM condition keys for agent-only policies, CloudWatch/CloudTrail auditing) with evaluated agent skills (npx skills add aws/agent-toolkit-for-aws/skills) and rules files; also works with Kiro, fx and other MCP clients or via aws configure agent-toolkit; successor to the AWS Labs MCP servers |
| Bifrost | Maxim AI | Open-source (Apache-2.0) Go AI gateway (npx -y @maximhq/bifrost or Docker) with a web UI that puts 23+ providers (OpenAI, Anthropic, Bedrock, Vertex, Azure, Ollama, Groq and more) behind one OpenAI-compatible API, with automatic failover, load balancing, semantic caching, an MCP gateway, virtual keys with budgets and rate limits, and Prometheus/tracing observability; documented setups route Claude Code, Codex CLI, Cursor, Gemini CLI, GitHub Copilot and OpenCode through it; ~11 µs overhead at 5k RPS in published benchmarks; enterprise tier adds clustering, guardrails and SSO; GitHub |
| neckbeard | bealesh | Open-source (Apache-2.0) DevOps/platform plugin for coding agents (/plugin install neckbeard@neckbeard in Claude Code; SKILL.md installs for Codex and Cursor) that turns an app repo into reviewable cloud infrastructure with the monthly bill attached: /neckbeard:analyze profiles the repo with file-and-line evidence and asks the questions code can't answer, /neckbeard:estimate prices the plan with Infracost before anything exists, and /neckbeard:sync renders OpenTofu, CI pipelines and Kubernetes manifests from a versioned, CI-tested module catalog via a deterministic Go CLI and hash-verified blueprint.yaml: the agent never invents modules; covers AWS, GCP and Azure (Cloud Run, ECS Fargate, Container Apps, EKS, GKE, AKS) with GitHub Actions or GitLab CI, validated with OpenTofu and Checkov |
| billion-context | ranxianglei | Open-source (MIT with an attribution term) context-compression proxy and plugin for coding agents (npm install -g billion-context, then bili claude, bili codex, bili opencode, bili pi or bili plugin install <host>) that sits between the agent and the Anthropic/OpenAI-compatible API and lets the model fold consumed conversation into layered, reversible summaries via injected compress, decompress, search_context and acp_status tools: prefix-cache friendly, aimed at ~5× fewer tokens and month-long single sessions inside a 100K window; supports Claude Code, Codex, OpenCode, pi, Gemini CLI, Qwen Code, Kimi, Hermes, Aider, GitHub Copilot CLI, Amp and more |
| Coucou | Louis Raillé (Louis-CFM) | Open-source (MIT) notch companion for AI coding agents on macOS (top-of-screen on Windows and Linux) with iPhone and Android apps: watches Claude Code, Codex, Cursor, Gemini CLI, Antigravity, Copilot CLI, OpenCode, Amp, Hermes and other sessions live (reads, edits, commands), lets you Allow/Deny permission requests and answer AskUserQuestion prompts from the notch, Lock Screen or Dynamic Island, shows Claude and Codex plan usage, and adds multi-provider chat (own keys or Ollama/LM Studio); 0.4 adds offline hold-to-talk dictation and call notes; GitHub |
| LazySkills | Alvin (alvinunreal) | Open-source (MIT) Go terminal UI and CLI for managing agent skills: one view of project, global, universal and agent-specific skills with per-agent visibility checks for Claude Code, OpenCode, Codex, Cursor, Gemini CLI and more; flags missing SKILL.md, invalid frontmatter, broken symlinks, missing lock entries and ghost skills, previews every command before safe reinstall/update/remove/prune actions, and searches and installs from skills.sh; lazyskills doctor --json gives a read-only health report and lazyskills restore rebuilds skills from lock files; install via the one-line script at https://lazyskills.sh/install, Homebrew (brew install --cask alvinunreal/tap/lazyskills) or go install; GitHub |
| Otari | Mozilla.ai | Open-source (Apache-2.0) OpenAI- and Anthropic-compatible LLM gateway you run yourself (Docker/Compose, Python 3.13+), built on Mozilla.ai's any-llm: one endpoint for 40+ providers with revocable virtual keys scoped by user, workspace and model, pre-dispatch budget checks, usage tracking, failover/weighted routing and an admin dashboard, plus optional code execution, web search, MCP and inference guardrails; documented setup routes Claude Code through it, and the standalone otari CLI (brew install mozilla-ai/tap/otari) adds otari hook Agent Guardrails: repo-committed .otari/ gates that refuse forbidden coding-agent tool calls and block turns ending in a forbidden diff, also checkable in CI; standalone, hybrid or hosted (otari.ai) modes; GitHub |
| skills.sh | Vercel Labs | Open-source (MIT) CLI for the open agent skills ecosystem (npx skills add vercel-labs/agent-skills) that installs SKILL.md skills from GitHub, GitLab, Azure Repos, any Git URL or a local path into OpenCode, Claude Code, Codex, Cursor and 75+ other agents, symlinking each agent to one canonical copy; skills find, list, update, remove, init and use (generate a prompt or start an agent with a skill without installing), private-repo support via existing Git/GitHub CLI auth, and a public skills.sh directory and leaderboard; GitHub |
| GitHub MCP Server | GitHub | Official open-source (MIT) MCP server that connects AI agents and IDEs to GitHub: browse and search code, manage issues and pull requests, monitor GitHub Actions runs and build failures, manage releases, and read code scanning and Dependabot alerts; use the hosted remote server (https://api.githubcopilot.com/mcp/, OAuth or PAT) from VS Code, Cursor, Claude, Windsurf and other MCP hosts, or run it locally via the ghcr.io/github/github-mcp-server Docker image or Go binary, with configurable toolsets, read-only mode and an insiders channel |
| TencentDB Agent Memory | Tencent Cloud | Open-source (MIT) team-level memory hub for AI agents: a Proxy you point the agent's base URL at (no plugin, hook or MCP server required) shares one memory server across Claude Code, Codex, CodeBuddy, WorkBuddy, DeepSeek Harness, Hermes and OpenClaw; distills conversations into layered Chat Memory, extracts reusable versioned Skills, turns docs into a linked Wiki and code into a CodeGraph, and adds team roles and review before personal assets are shared; self-hosted with ./start-all.sh (memory-core + memory-hub + proxy) and a web panel |
| Phone Harness | Shawn Pana | Open-source (MIT) Python CLI and agent skill that lets Claude Code, Codex, Hermes or any coding agent drive a real phone: iPhone through macOS iPhone Mirroring, Android over adb (macOS, Linux, Windows), or a hosted Phone Harness Cloud device; helpers such as screen(), find_text, tap_text, type_text and open_app let the agent see the screen, tap, type and read results with no jailbreak, Xcode or on-phone install; phone-harness doctor checks setup, and cloud Android is billed per minute after 100 free minutes; GitHub |
| Harness2 | Google Research | Open-source (Apache-2.0) Python framework for recursive agent harnessing (Harness²): with model weights fixed, it refines the harness around a coding agent (system prompt, guardrails, memory, tool guidance, subagents, skills, scripts and plugins) per task through a contrastive Propose-Probe-Compose loop, then carries adopted harnesses and evidence-backed lessons into a persistent improvement harness for future tasks; parallel and sequential recursion, OpenCode and Codex executors, harness2-demo runs locally without model credentials, and JobBench, LAB and WorkBuddy-Bench runners (harness2-setup, harness2, harness2-judge, harness2-report) use Vertex AI models; Paper |
| AgentsView | Kenn Software | Open-source (MIT) local-first session archive and analytics for coding agents: discovers Claude Code, Codex, Cursor, Copilot CLI, Gemini CLI, Antigravity, Devin CLI, Junie, Goose, Aider and 20+ other agents' sessions, syncs them into local SQLite and serves a web UI and CLI (agentsview serve, agentsview usage daily) for full-text session search, transcripts with tool errors, per-session stats and token and cost tracking; desktop app for macOS/Windows (brew install --cask agentsview), install script, or Docker image, with optional PostgreSQL, ClickHouse and DuckDB sync for teams; GitHub |
| Laminar | Laminar (YC S24) | Open-source (Apache-2.0) observability platform for AI agents, written in Rust: OpenTelemetry-native tracing SDK that auto-instruments Vercel AI SDK, LangChain, Browser Use, Stagehand, OpenAI, Anthropic and Gemini; plain-English Signals (e.g. "agent is stuck in a loop") scored on every run with Slack alerts, powered by its RL-trained flow-1 trace-error model; evals SDK/CLI for local or CI runs, SQL dashboards, datasets, and an MCP server and CLI so coding agents can query traces and debug failures; self-host with Docker Compose or use the managed cloud (Free, Starter $30/mo, Pro $150/mo, Enterprise); GitHub |
| ThreadKeep | Interesting Concepts LLC | ThreadKeep helps developers carry project context between AI coding sessions through structured MCP/HTTP checkpoints and editable context packs. Optional AI context updates turn conversations into reusable context; checkpoint saves and reads themselves do not call an AI model. |
| System One API | Ziyu Li / System One | Open AI API gateway with a unified TypeScript SDK (@system-one-ai/core). One interface to route across models for agent and developer workflows. GitHub https://github.com/ziyu/system-one-sdk currently has about 2 stars — disclosing honestly.; GitHub |
| Hallmark | Together AI | Open-source design skill for Claude Code, Codex, Cursor, OpenCode and GitHub Copilot that "refuses to look AI-generated" (npx skills add nutlope/hallmark -g or as a Claude Code plugin): deals three of 48 themes, builds pages around their job, and lints output in a real browser at eight widths with sloplint against tiered gates (Floor, Reflex, Finish) before a fresh-context reviewer signs off; /hallmark redesign and /hallmark audit work on existing pages; GitHub |
| Codync | leepokai | Free, open-source (Apache-2.0) "Grok Bot alternative" that turns the coding agents on your computer (Claude Code, Codex, Cursor, Gemini, Copilot, OpenCode, Pi and ~40 more via the Agent Client Protocol registry) into persistent named bots you message from a native iPhone app, an Electron desktop app for macOS/Windows/Linux or a terminal UI; group chats, bot-to-bot help via a built-in team MCP server, approval cards, push notifications and remote screen, with a Rust codync-host and an end-to-end encrypted free relay; GitHub |
| Respan Gateway | Respan (YC W24) | Open-source (MIT or Apache-2.0) Rust AI gateway (pip install respan-gateway or cargo install respan-gateway) that lets OpenAI Chat Completions/Responses, Anthropic Messages and Gemini clients call any model, with streaming, tools and reasoning translated; puts MCP servers and built-in tools behind chat requests, adds routing, fallback, retries and hedged requests, a plugin system (guardrails, PII masking, pricing), and a live request log with cost and latency at localhost:8080/ui; GitHub |
| Android Skills | Official open-source (Apache-2.0) library of 30+ Agent Skills (SKILL.md) from the Android team that ground coding agents in developer.android.com guidance where evaluations show LLMs underperform: AGP 9 upgrades, XML-to-Jetpack Compose migration, Navigation 3, edge-to-edge, R8 analysis, ANR audits, CameraX, Media3 Cast, Play Billing upgrades, Credential Manager, intent and permission security, Wear OS, TV, cars, XR and on-device AI (LiteRT, ML Kit GenAI); install with android skills add from Android CLI for every detected agent, or as a Claude Code / Codex plugin marketplace; GitHub |
|
| Entire | Entire | Git forge built for agent–human collaboration from former GitHub CEO Thomas Dohmke's startup, driven by an open-source (MIT) entire CLI: entire enable installs hooks for Claude Code, Codex, Cursor, Copilot CLI, OpenCode, Antigravity, Factory Droid, Pi and more so every commit with agent work gets a checkpoint (prompts, transcript, files touched) stored as git refs, searchable with hybrid entire search; host repos on Entire or mirror GitHub ones (pushes still go to GitHub) across regions, and review changes as Trails, a heads-up display scored on every push by per-repo eval runners (risk, drift, security, review) instead of a raw diff; optional entire-graph code map and entire-brain project memory, plugins; checkpoint capture needs no account, hosted use is billed for agents with a $100 signup credit; GitHub |
| Selan | Selan UAB (submitted by the Selan team with AI assistance) | Gateway for teams running Claude Code and Codex on company accounts: logs who ran each request and what it cost, sets per-person spend limits and an effort ceiling per model maker, and keeps provider keys off laptops and out of CI secrets (CI jobs hold only a Selan token). Optional secret redaction on request bodies before they reach the model, off until an owner enables it. Free for up to 3 seats, paid plans from EUR 29/month for up to 10 seats (paid plans currently for EU/EEA organisations), and model usage bills to your own provider accounts. Submitted by the Selan team with AI assistance. |
| Expo Skills | Expo | Official open-source (MIT) library of 20+ Agent Skills (SKILL.md) from the Expo team for building, deploying, upgrading and debugging Expo and React Native apps: an expo-overview router skill, free framework skills (Expo Router, animations, native UI, @expo/ui, data fetching, DOM components, web-to-native migration, native modules, brownfield, SDK upgrades) and clearly labelled paid EAS skills (app-store builds and submission, Hosting, Workflows, Update, Observe, remote simulators); install with npx skills add expo/skills, as the expo plugin in the official Claude Code and OpenAI-curated Codex marketplaces, or for Cursor, Copilot, Gemini, OpenCode and other agents, and pair with the Expo MCP server; GitHub |
| Sandbox Runtime (srt) | Anthropic | Open-source (Apache-2.0) research-preview sandbox developed for Claude Code that enforces filesystem, network and Unix-socket restrictions on agents, local MCP servers, bash commands or any process at the OS level without a container: npm install -g @anthropic-ai/sandbox-runtime, then wrap commands with srt (or use it as a library); deny-by-default network allowlists via built-in HTTP and SOCKS5 proxies, read/write path rules in ~/.srt-settings.json, using sandbox-exec on macOS, bubblewrap on Linux and a dedicated user plus Windows Filtering Platform rules on Windows, with live violation monitoring on macOS |
| cc-skills-golang | Samuel Berthe (samber) | Open-source (MIT) collection of 40+ atomic, cross-referencing Agent Skills for production Go projects from the author of samber/lo, bootstrapped from real Go commits and human-reviewed: code style, naming, error handling, safety, concurrency, testing, benchmarks, performance, observability, security, modernize, refactoring, project layout and CI, plus library skills for gRPC, GraphQL, Cobra, Viper, Wire, Fx, testify and samber/* packages; published token weights and error-rate gaps per skill; install with npx skills add samber/cc-skills-golang --all, as a Claude Code plugin, Gemini CLI extension, or for Cursor, Copilot, Codex, OpenCode and Antigravity |
| Sqwakie | Alex Carney | A walkie-talkie for your team's AI agents: one live dashboard and peer-to-peer messaging with an encrypted relay fallback (CLI or MCP) for Claude Code and Codex agents across a team's machines, with project kanbans agents can start, review, and finish. Free for 2 people; paid per-person plans. Source-available under FSL-1.1-ALv2 (not OSI open source). Disclosure: I make Sqwakie. Support via GitHub issues on the walkie repo ([email protected] has no MX yet).; GitHub |
| Skill Recorder | Microsoft | Open-source (MIT) desktop app for macOS and Windows 11 that records one real work session (screen activity, app and window switches, visited pages, an optional recorded terminal and spoken narration) and uses the bundled GitHub Copilot CLI to reconstruct it as an editable intent plus ordered steps, then generates a reusable SKILL.md Skill or a scheduled/triggered Automation that prefers native agent tools (like gh or web_fetch) over replaying clicks; skills can be reviewed, exported or added to Microsoft Autopilots such as Scout, Copilot Cowork and Copilot Studio; requires a GitHub account with Copilot access and installs as a commit-pinned source release |
| MCode Stack | MiniMax | Open-source (MIT) collection of the workflows, skills and checklists the MiniMax Code team uses to build MiniMax Code, packaged as per-engineer stacks (ronny, jiachen) of SKILL.md skills such as subagent-split-check, tool-budget-audit, context-tax-audit, long-session-check, handoff-doc, long-command-check and independent-review, each paired with an MCode Engineering Notes article; installs as a plugin marketplace in MiniMax Code, Claude Code (claude plugin install ronny --marketplace MiniMax-AI/mcode-stack) and Codex, or by copying single skills into ~/.agents/skills/ |
🖊️ Editor Platforms with Native AI Features
Established editors that have shipped first-party AI capabilities.
| Tool | Company | Notes |
|---|---|---|
| Visual Studio Code | Microsoft | Copilot built-in + massive extension ecosystem for third-party AI tools |
| Xcode | Apple | Predictive code completion + Swift Assist for Apple platform development |
| JetBrains IDEs | JetBrains | IntelliJ, PyCharm, WebStorm etc. with JetBrains AI (Junie agent) built in |
| Visual Studio | Microsoft | GitHub Copilot deeply integrated; .NET and C++ focused |
🔧 Hardware & Edge AI
Agentic tools and runtimes that bring AI to hardware design and local/edge inference.
| Tool | Company | Notes |
|---|---|---|
| Copperhead | Copperhead | "Cursor for circuit boards": agentic CLI that designs, documents, and validates real KiCad PCBs from a prompt; runs inside Claude Code, Cursor, and Codex (GitHub) |
| Personal AI Router | NVIDIA | Local inference router (PAIR) that distributes requests across machines on a home network; exposes Ollama- and OpenAI-compatible endpoints for local agents and multi-agent workloads |
| WebLLM | MLC AI | High-performance in-browser LLM inference engine accelerated by WebGPU with full OpenAI API compatibility: run open models locally in the browser with no server |
| Magnitude | Magnitude (YC) | Open-source (Apache-2.0) local inference engine and desktop app (macOS/Windows/Linux) that profiles your hardware, estimates tok/s for every open model before download, recommends and tunes the best fit (speculative decoding, placement, batching) on Apple Silicon, NVIDIA, AMD or CPU-only, then connects it in one click to Pi, OpenCode, Hermes, OpenClaw, Codex, Claude Code, Oh My Pi or Cline; models load on demand and unload when idle, fully offline; bundles the magnitude CLI; GitHub |
| text-to-cad | earthtojake | Open-source (MIT) plugin and skills that give coding agents local CAD workflows: generate 3D models as STEP, GLB, STL or 3MF, run design-for-manufacturing checks, produce engineering drawings and connect to 3D-printing, sheet-metal and CNC fabrication services; a bundled cadgen mcp server and CAD Viewer render models in chat or the browser, with plugins for Claude Code, Claude Desktop and Codex and skills for Cursor, Gemini, Grok and other agents; GitHub |
Last updated: July 2026
[!WARNING]
Disclaimer: This directory is manually curated and may contain outdated information regarding pricing, features, or availability. Please verify details directly on the vendor's website for accuracy.
[!NOTE]
Some links in this guide are referral links (e.g., Emergent.sh). Using them helps support the ongoing maintenance of this curated list at no additional cost to you.
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found