🦗 MantisClaw - Desktop CUA (Computer Use Agent)

---

Multi-Platform - Windows / Linux / Mac OS - Intel / Mac OS Silicon

---

Use Computer & Browser

• Open webcam - Use Real Browser (Pass Cloudflare & bot verifications) -Take Screenshot - scroll - click_button - login - check-email

---

Unlimited Agents

---

• Unlimited Scenarios - Unlimited Skills (It writes, debug, executes skills by itself -Self-healing)

Channels

Connect and manage multiple messaging channels — WhatsApp, Telegram, Slack, Built-in Chat, Blender Chat .. and more.

Multi-Platform Autonomous Agent — Control from WhatsApp, Telegram, Slack, and more

The best GUI Computer Use Agent. No complex setup — just download and run. All features are already built-in.

🌐 Website  ·  ⬇️ Download  ·  🎬 Demo Videos  ·  📸 Screenshots

---

MantisClaw integrates autonomous agents across multiple channels so your team can command, automate, and supervise real operations from chat. Use direct or group conversations to run tasks, trigger workflows, and keep execution secure with local-first data control.

🎬 Demo Videos

▶ Demo Video 1	▶ Demo Video 2
▶ Demo Video 3	▶ Demo Video 4

---

🚀 What's New

• WhatsApp, Telegram, Slack, Built-in ChatUI channels
• MantisClaw Blender MCP Control UI
• MCP support for extensible tool discovery
• OpenAI-compatible Tools & Advanced HTTP Call Support
• Workflow — POC (under development)
• 50+ Node Registry POC (under development)
• Templates POC (under development)
• Bug Fixes & Stability Improvements

---

🔒 Why MantisClaw Stands Out

Feature	Details
Local Database	Your accounts & keys never leave your computer
Runtime Key Injection	Keys injected at runtime, never sent to LLM
Built-in Office Engine	Word, Excel, PowerPoint & PDF generation
Real Browser	The only Agent that can use a real browser
Multi-Platform CUA	First & only GUI Multi-Platform CUA Agent
Token Efficient	Uses less tokens than competitors
Self-Healing Code	Writes code, debugs, fixes & self-heals automatically
Auto Skills	Creates skills & skill settings by itself
Unlimited Agents	Unlimited Agents & unlimited Scenarios (SOUL)
Account Security	Sub-accounts for granular access control
Completely Free	All features included at no cost

---

⚙️ Platform Capabilities

Autonomous Agent Integrated with Multiple Channels

Run an autonomous agent through chat conversations across channels and keep operations inside real communication flows.

Control Your Agent from Any Channel

Start tasks, steer behavior, and trigger workflows by messaging your agent from your preferred channel.

Direct + Group Command Support

Use both direct chats and groups as command surfaces for agent actions and automation routines.

Code and Task Execution

Ask the agent to write code, run scripts, call APIs, and produce business outputs from channel instructions.

Secure Runtime Settings

Sensitive values are injected at runtime and kept out of chat prompts whenever possible.

Local-First Agent Memory

Portable PostgreSQL keeps your operational data local unless you explicitly route data to an external model provider.

---

🔄 How MantisClaw Works

From incoming channel message to autonomous action — here is what happens under the hood.

Step	Phase	Description
1	Message Arrives	A message from a connected channel reaches MantisClaw. The system identifies the sender, resolves the target agent mapping, and queues the message for processing.
2	Context Assembly	The last 20 messages are loaded from chat history plus the agent's persona/scenario. This context window is sent alongside all available tools to the LLM.
3	Tool-Call Loop (25 Turns)	The AI decides actions in a single loop: run Python code, browse the web, execute skills, create files, or call APIs. Each tool result feeds back into the next decision.
4	Self-Healing Execution	If Python or a skill fails, the AI reads the traceback, fixes the code, auto-installs missing packages, and retries — up to 3 attempts before reporting.
5	Output Delivery	Text responses go back to the source channel. Generated files (Excel, Word, PDF, images) are auto-detected and sent as attachments to the original chat.
6	Learn & Schedule	Skills discovered during execution are saved for reuse. Scheduled tasks run autonomously on cron and deliver results directly to connected channels.

---

🏗️ Agent Architecture

Component	Description
Go + Wails	Desktop runtime with embedded web server and system tray
Embedded PostgreSQL	Local-first database — no external server required
Playwright + Chromium	Persistent browser context with cookies across restarts
Python Kernel	Embedded interpreter with auto pip-install and SETTINGS injection

---

🚦 Agent Control Flow

1. Create your account and complete verification.
2. Connect your channels securely.
3. Choose direct chats and groups as control channels.
4. Assign those channels to your autonomous agent.
5. Control and monitor agent actions directly from your channels.

---

❓ Why MantisClaw?

MantisClaw helps teams manage one autonomous agent across multiple messaging channels. Send commands from chat, monitor every execution step, and receive outcomes in the same conversation flow.

Pillar	Description
Multi-Channel Control	Control one agent from WhatsApp, Telegram, Slack, and other communication channels.
Local and Secure Execution	Sensitive data stays local, while runtime settings keep workflows secure, traceable, and auditable.
Automation and Productivity	Combine coding, API calls, file generation, and scheduled tasks in one agent to accelerate operations.

---

⚡ See it in Action

Watch how a single channel message flows through the MantisClaw engine — from ingestion through the AI tool-call loop to autonomous execution and response delivery.

Feature	Details
Single Tool-Call Loop	25 turns max — the AI sees all tools at once and picks the right action each turn. No multi-stage overhead.
Self-Healing Pipeline	Python errors auto-trigger fix → retry cycles. Missing pip packages get installed on the fly.
Scheduled Autonomy	Cron tasks feed directly into the AI Brain. Results are auto-delivered to your channels on schedule.

---

📸 Platform Screenshots

---

Channels

Connect and manage multiple messaging channels — WhatsApp, Telegram, Slack, and more.

---

Accounts

Manage all your connected accounts, credentials, and linked services in one centralized dashboard.

---

Advanced MCP

Configure advanced Model Context Protocol settings for fine-grained control over AI model interactions.

---

Advanced Tools

Access powerful advanced tools including code execution, API testing, and debugging utilities.

---

Advanced Workflows

Design complex multi-step workflows with conditional logic, branching, and parallel execution paths.

---

Agents

Create and configure autonomous agents with custom personas, scenarios, and channel assignments.

---

AI-Assisted Node Registry

Browse and register workflow nodes with AI-powered suggestions and auto-configuration.

---

Architecture

Visual overview of the MantisClaw system architecture showing all integrated components.

---

Blender Control via MCP

Control Blender 3D directly from chat using MCP integration for 3D modeling and rendering tasks.

---

Browser Playbooks

Record and replay browser automation sequences as reusable playbooks for web tasks.

---

Built-in Chat

Use the built-in chat interface to interact with your agent directly from the MantisClaw desktop app.

---

LLM Settings

Configure language model providers, API keys, temperature, token limits, and model selection.

---

MantisClaw + Blender

See MantisClaw controlling Blender to create 3D scenes and render outputs from chat commands.

---

Blender Integration Demo

Step-by-step demonstration of Blender 3D integration with MantisClaw agent control.

---

MCP Configuration

Set up Model Context Protocol servers and tools to extend your agent capabilities.

---

Node Registry

Browse available workflow nodes, skills, and integrations in the central registry.

---

Reset Option

Safely reset agent state, clear conversation history, or restore default configurations.

---

Scenarios

Define agent scenarios with system prompts, behavior rules, and operational boundaries.

---

Scheduled Tasks

Set up cron-based scheduled tasks that run autonomously and deliver results to your channels.

---

Skills

Manage reusable skills — code snippets, API integrations, and automation routines your agent can invoke.

---

Skill Settings

Configure skill parameters, runtime settings, and execution preferences for each registered skill.

---

Tool MCP

View and manage MCP tool definitions available to your agent during execution.

---

Workflow List

Browse all created workflows with status indicators, execution history, and quick actions.

---

Workflows

Visual workflow editor to build, connect, and orchestrate automation pipelines with drag-and-drop.

---

💼 Use Cases Across Teams

Use Case	Description
Daily Task Assistant	Automate morning checklists, prepare end-of-day summaries, and send daily reminders in team channels.
Routine Task Automation	Handle repetitive actions like status collection, recurring reports, and regular follow-up pings.
Workflow Orchestration	Run multi-step workflows across tools: collect input, execute logic, generate outputs, and post results automatically.
Dynamic API and Tool Calls	Trigger dynamic calls to external APIs, internal services, or scripts based on natural language instructions.
Customer Support	Create ticket summaries, suggest replies, and push escalations to CRM tools.
Sales Operations	Qualify leads from chat, prepare follow-up drafts, and update pipeline stages automatically.
Marketing Workflows	Draft campaign copy, collect feedback from group chats, and generate content variants.
Finance Requests	Prepare weekly KPI snapshots, reconcile simple reports, and send approval-ready summaries.
Engineering Automation	Run scripts, inspect logs, create troubleshooting notes, and deliver patch suggestions.
DevOps Coordination	Trigger operational checks, post incident updates, and automate recurring maintenance routines.
HR and People Ops	Answer policy questions, automate onboarding reminders, and collect status confirmations.
Project Management	Convert chat decisions into task lists, sync action items, and publish progress digests.
Executive Reporting	Aggregate updates from multiple teams and produce concise daily or weekly briefings.
Procurement and Operations	Track purchase requests, route approvals, and notify stakeholders when vendor milestones are reached.
Compliance and Audit Trails	Log key decisions, preserve execution history, and generate audit-ready summaries for internal controls.
Knowledge Operations	Turn chat outcomes into reusable playbooks, FAQs, and process notes for continuous team learning.
Word Documents and Letters	Generate formal letters, proposals, contracts, and branded Word documents directly from chat prompts.
Excel Sheets and Reports	Create structured spreadsheets, formulas, budget trackers, KPI tables, and operational report templates.
PowerPoint Presentations	Prepare meeting decks, investor summaries, training slides, and visual storylines ready for presentation.
PDF Documents	Export polished PDFs such as invoices, summaries, compliance packs, and client-ready documentation.

---

🎬 Demo Videos

▶ Demo Video 1	▶ Demo Video 2
▶ Demo Video 3	▶ Demo Video 4

---

⬇️ Download

Available for Windows, Linux, Apple Intel, and Apple Silicon.

Platform	Download Link
Windows (amd64)	MantisClaw_win_amd64.zip
Linux (amd64)	MantisClaw_linux_amd64.zip
macOS Intel (amd64)	MantisClaw-MacOs_Intel_amd64.zip
macOS Silicon (arm64)	MantisClaw-MacOs_Silicon_arm64.zip

Visit mantisclaw.yapzek.ai for setup instructions, compatibility notes, and latest version details.

---

🔧 Comprehensive Feature Summary

Component	Features	Scope
Communication	WhatsApp, Telegram, Slack, Web Chat	4 channels
Agents	My Agents, Desktop Agents, Browser, Skills	4 types
Workflows	DAG execution, 50+ node types, versioning	Full
Code Execution	Python kernel, sandbox, auto-packages	Full
Blender	Scene info, objects, materials, asset import	Full
Database	PostgreSQL, 25+ tables, vector memory planned	Full
LLM	8 providers, multi-model support	Full
Tools	Browser (10+), Skills (custom), HTTP, MCP, Workflow	50+ tools
Self-Healing	Skill code fixes, tool schema recovery	Full
Scheduling	Cron jobs, prompt/skill/pipeline types	Full
Frontend	22+ pages, React Flow DAG editor	Full
Playbooks	JSONL format, recording, LLM vision, 15+ actions	Full
Node Registry	50+ builtin, custom creation, AI generation	Full
Desktop	Wails, system tray, 3 runtimes embedded	Full
Security	JWT auth, email/SMS verification, sandbox	Full
History	Chat logs, runs, playbooks, task history, vector search planned	Full

Key Metrics

• API Endpoints: 100+
• Database Tables: 25+
• Frontend Routes: 22
• LLM Providers Supported: 8
• Tool Categories: 5
• Workflow Node Types: 50+
• Browser Tool Functions: 10+
• Max Concurrent Workflows: 10
• Desktop Agents Per User: Unlimited

---

🖥️ Cross-Platform Support

OS	Status	System Tray	Notes
Windows	✅ Full	✅	All features, system tray working
macOS	✅ Partial	❌	Wails native menus, no GTK conflict
Linux	✅ Partial	❌	GTK signal handler issue with Wails

---

🔗 Links

• 🌐 Website — mantisclaw.yapzek.ai
• 📊 MantisClaw vs Competitors
• ⚔️ OpenClaw vs MantisClaw vs NemoClaw
• 📜 Terms of Service
• 🔐 Privacy Policy

---

© 2026 MantisClaw — mantisclaw.yapzek.ai