Skillget

chatml

agent
Guvenlik Denetimi
Uyari
Health Gecti
  • License — License: NOASSERTION
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 39 GitHub stars
Code Uyari
  • process.env — Environment variable access in .github/workflows/release.yml
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This tool is a native macOS application that orchestrates multiple AI coding agents to run in parallel. It allows developers to assign distinct coding tasks to AI agents, which then independently write code, run commands, and create PRs within isolated git worktrees.

Security Assessment
Overall Risk: Medium. The application provides Claude with full agent autonomy, meaning it explicitly executes shell commands, reads, and writes files within your local environment. It also makes external network requests to the Anthropic API to function, requiring you to provide an API key. While the tool itself does not request dangerous system permissions outside of macOS, the agent inherently has broad access to the local repositories it is assigned. No hardcoded secrets were detected in the codebase, though environment variables are accessed during its release workflow.

Quality Assessment
The project appears to be an active and maintained tool, with its latest repository push occurring today. It is backed by a modest but positive community response with 39 GitHub stars. The codebase includes continuous integration workflows, demonstrating good development practices. While the README specifies a GPL-3.0 license, the automated repository metadata lists the license as "NOASSERTION," which is a minor administrative discrepancy that might require verification for enterprise use.

Verdict
Use with caution — while the orchestrator itself is safe, the underlying AI agents are given high autonomy to execute commands and modify files on your local system.
SUMMARY

AI Agent Orchestrator for Claude Code

README.md

ChatML mascot

ChatML

Run AI coding agents in parallel. Each task gets its own worktree.

Release License CI Platform Follow @chatmlx

ChatML — AI coding agents running in parallel with isolated worktrees

ChatML is a native macOS app for AI-assisted development. Instead of chatting with an AI and copy-pasting code, you describe a task and Claude does the work — reading files, writing code, running commands, and creating PRs — all inside an isolated git worktree so nothing touches your working directory.

Spin up five sessions at once: one refactoring auth, one adding an API endpoint, one fixing a bug, one writing tests, one reviewing a PR. They all run in parallel without stepping on each other.

How It Works

1. Add a repo          →  Register any git repository as a workspace
2. Create a session    →  ChatML creates an isolated worktree + branch
3. Describe the task   →  Claude writes, edits, and runs code directly
4. Review & merge      →  Check the diff, open a PR, done

Sessions are fully isolated. Each one gets its own branch, its own working directory, and its own AI agent. No merge conflicts between parallel tasks.

Download

Download the latest release (.dmg for macOS)

Or build from source:

git clone https://github.com/chatml/chatml.git && cd chatml && make dev

You'll need an Anthropic API key — the onboarding wizard will walk you through setup.

Key Features

Worktree-Isolated Sessions

Every session creates a git worktree with its own branch and working directory. Sessions are fully isolated — no merge conflicts between parallel tasks, no accidentally committing to the wrong branch.

Full Agent Autonomy

Powered by the Claude Agent SDK. Claude has full tool access within each session: reading and writing files, running terminal commands, searching the codebase, and executing multi-step workflows. Watch everything happen in real time — text streams as it's generated, tool calls show their status and duration, sub-agents are tracked independently.

Built-In Code Review

Start a review conversation and Claude examines the session's changes with inline comments, severity levels, and resolution tracking.

Pull Requests & CI/CD

Push a branch and create a GitHub PR without leaving the app. Claude generates PR descriptions from the diff. Live status polling tracks checks, merge conflicts, and review state. View GitHub Actions runs, drill into job logs, and let Claude analyze failures.

Skills Marketplace

25+ specialized prompt templates: TDD guidance, security audits, systematic debugging, architecture decision records, and more. Install per-session, or create your own.

Budget & Context Controls

Set cost limits, turn limits, and thinking token budgets. Monitor context window utilization and token usage in real time. Automatic git stash-based checkpoints let you revert to any previous state.

All Features
  • File browser & editor — Session-scoped tabs, syntax highlighting, side-by-side diffs, direct editing
  • Terminal integration — Full PTY terminal emulation, up to 5 terminals per session
  • Extended thinking & plan mode — Control how deeply Claude reasons before acting
  • Linear integration — OAuth-based issue discovery, context, and management
  • MCP support — Built-in server + custom MCP server configuration (stdio, SSE, HTTP)
  • Session management — Priority, status tracking, pinning, archiving with AI-generated summaries
  • Branch sync — Detect when behind origin/main, sync via rebase or merge
  • Checkpointing & rewind — Automatic checkpoints, revert files to any previous state
  • Keyboard shortcuts — 30+ shortcuts for fast navigation
  • Auto-update — Automatic update detection and installation

How It Compares

ChatML Cursor / Windsurf Claude Code CLI GitHub Copilot
Parallel tasks Each task in its own worktree Single workspace Single terminal session Single file context
Agent autonomy Full tool access: file I/O, terminal, git Editor-integrated suggestions Full tool access Inline completions
Code review Built-in with inline comments External External External
CI/CD monitoring GitHub Actions integration External External External
PR workflow Create, track, merge natively External External GitHub integration
Open source GPL-3.0 Proprietary Proprietary Proprietary

Architecture

ChatML is a polyglot app with four layers:

┌─────────────────────────────────────────────┐
│              Tauri 2 (Rust)                 │
│         Native macOS Desktop Shell          │
│                                             │
│  ┌───────────────┐  ┌───────────────────┐   │
│  │  Next.js 15   │  │   Go Backend      │   │
│  │  React 19     │◄─►  REST + WebSocket │   │
│  │  Tailwind 4   │  │  SQLite           │   │
│  │  Zustand      │  │  Git Operations   │   │
│  └───────────────┘  └────────┬──────────┘   │
│                              │              │
│                     ┌────────▼──────────┐   │
│                     │  Agent Runner     │   │
│                     │  Node.js + Claude │   │
│                     │  Agent SDK        │   │
│                     └───────────────────┘   │
└─────────────────────────────────────────────┘
Layer Tech Role
Desktop Shell Tauri 2, Rust Native window, menus, PTY, secure storage, auto-update
Frontend Next.js 15, React 19, Tailwind CSS 4 UI, state management, WebSocket client
Backend Go, SQLite REST API, WebSocket, git operations, agent lifecycle
Agent Runner Node.js, Claude Agent SDK AI agent processes, tool execution, streaming

For deeper technical details, see the architecture documentation.

Contributing

We welcome contributions! Whether it's bug fixes, new features, documentation, or design work — there's plenty to do.

# Fork the repo, then:
git clone https://github.com/YOUR_USERNAME/chatml.git
cd chatml
make dev

See CONTRIBUTING.md for the full development setup, OAuth configuration, and PR process.

Areas Where Help is Needed

  • Cross-platform support — Linux and Windows builds
  • Testing — Frontend component tests, E2E tests
  • Multi-model support — Providers beyond Claude
  • Documentation — Guides, tutorials, API docs
  • UI/UX — Design improvements, accessibility
  • Agent capabilities — New skills, MCP integrations

Looking for a place to start? Check good first issues.

CLA Requirement

We require all contributors to sign a Contributor License Agreement before we can accept contributions. This is handled automatically when you open your first PR.

Roadmap

  • Linux & Windows support — Cross-platform desktop builds
  • Multi-model support — Use different AI providers beyond Claude
  • Team collaboration — Shared workspaces and session handoff
  • Plugin system — Community-built skills and integrations
  • Self-hosted backend — Run the backend as a standalone server
  • Voice interaction — Talk to your agent

Want to help with any of these? Open an issue or jump into a discussion.

Security

API keys are stored encrypted with AES-256-GCM. Agent sessions run in isolated worktrees. No telemetry. All communication stays on localhost.

See SECURITY.md for our security policy and how to report vulnerabilities.

License

ChatML is licensed under the GNU General Public License v3.0. You are free to use, modify, and distribute this software, provided that derivative works are distributed under the same license.

⭐ If you find this interesting, please star the repo.

Built with obsessive attention to developer workflow.
Report a Bug · Request a Feature · Discussions · @chatmlx

Yorumlar (0)

Sonuc bulunamadi