ClawCrew - Multi-Agent AI Teams on OpenClaw

What is ClawCrew?

ClawCrew is a framework for creating multi-agent teams that collaborate on tasks. Instead of one AI doing everything, you define specialized agents with focused roles, and an orchestrator coordinates their work.

🧬

Each Agent Evolves

Train and tune each agent's capabilities separately — refine the designer's taste, sharpen the coder's style, improve the tester's coverage.

👔

You're the Manager

In Telegram, talk to the orchestrator and watch every agent work in real-time. Jump in anytime — give feedback, correct course, or let them run.

✨

Workspace = Capability

Each agent's capabilities are defined by its workspace — SOUL.md, skills, and tools. Change the workspace, change the agent.

Core Concept

┌──────────────────────────────────────────────────────────────┐ │ ClawCrew Framework │ ├──────────────────────────────────────────────────────────────┤ │ │ │ Orchestrator Agent (workspace_orca/) │ │ ├── SOUL.md → Personality & rules │ │ ├── AGENTS.md → Behavior guidelines │ │ └── skills/ → Available capabilities │ │ │ │ │ spawns │ │ ▼ │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │ │ │ Agent A │ │ Agent B │ │ Agent C │ │ │ │ workspace_A/ │ │ workspace_B/ │ │ workspace_C/ │ │ │ └──────────────┘ └──────────────┘ └──────────────┘ │ │ │ │ Each agent = unique workspace = unique capabilities │ │ │ └──────────────────────────────────────────────────────────────┘

Example Teams

💻 Software Dev Team

🦑 OrcaBot (Orchestrator) ├── 🎨 DesignBot → API design, specs ├── 💻 CodeBot → Implementation └── 🧪 TestBot → Testing, coverage

Automated code generation with quality gates

📚 Research Team

📚 ResearchLead (Orchestrator) ├── 🔍 SearchBot → Web research ├── 📊 AnalystBot → Data analysis └── ✍️ WriterBot → Report writing

Deep research with structured output

🎬 Content Team

🎬 ContentLead (Orchestrator) ├── 💡 IdeaBot → Brainstorming ├── ✏️ DraftBot → Writing drafts └── 🔍 EditorBot → Review, polish

Blog posts, marketing copy, documentation

🎯 Support Team

🎯 SupportLead (Orchestrator) ├── 🔍 TriageBot → Categorize issues ├── 💬 ReplyBot → Draft responses └── 📈 EscalateBot → Complex handling

Automated support ticket handling

Software Dev Team

The included example shows a complete software development crew with design, implementation, and testing phases.

Agent	Role	Deliverables
🦑 OrcaBot	Orchestrator	Task breakdown, quality gates, final delivery
🎨 DesignBot	Architect	API specs, types, edge cases
💻 CodeBot	Engineer	Implementation, docstrings
🧪 TestBot	QA	Unit tests, coverage report

Design Phase

OrcaBot spawns DesignBot → API spec created → OrcaBot reviews ✅

Code Phase

OrcaBot spawns CodeBot → Module implemented → OrcaBot reviews ✅

Test Phase

OrcaBot spawns TestBot → Tests written & run → OrcaBot reviews ✅

Final Delivery 🚀

Code + Tests + Documentation delivered with full coverage

Getting Started

Prerequisites: OpenClaw installed and configured. See OpenClaw docs.

⚡

One-Click Install

curl -sSL https://raw.githubusercontent.com/lanxindeng8/clawcrew/main/install.sh | bash

Auto-detects OS, installs dependencies, runs setup wizard.

📦

Install via pip

pip install clawcrew

Standard Python package installation.

🐳

Docker

git clone https://github.com/lanxindeng8/clawcrew
cd clawcrew
cp .env.example .env
docker-compose up -d

After Installation

# 1. Reload your shell (or restart terminal)
source ~/.profile   # or ~/.zshrc depending on your shell

# 2. Configure ClawCrew (Telegram bot, OpenClaw settings)
clawcrew init

# 3. Start the system
clawcrew start

# 4. Verify everything is running
clawcrew status

Send a Task

Telegram: Send a message in your Telegram group:

Create a Python module to calculate distance between two points

CLI: Chain multiple agents with automatic context passing:

clawcrew chain "Create a distance calculator module" design code test

OrcaBot will coordinate DesignBot → CodeBot → TestBot to complete the task.

CLI Reference

Command	Description
`clawcrew init`	Interactive setup wizard
`clawcrew start / stop / status`	Manage the OpenClaw gateway
`clawcrew run <agent> -t "task"`	Run a single agent
`clawcrew chain "task" design code test`	Chain multiple agents with auto context passing
`clawcrew agents`	List available agents
`clawcrew github analyze --url <repo>`	Analyze a GitHub repository

Roadmap

✅

v0.3 - Current

PyPI package, CLI tools, GitHub integration, Docker deployment, Repo Mode workflows, Dashboard

🔄

v0.4 - In Progress

Easy onboarding, Role templates, Shared context, Team memory, Workflow editor

🔮

Future

Multi-team communication, Code review workflows, CI/CD hooks, Organization policies

Recent Updates

NEW Phase 4: Installation & DX

PyPI Package — Install with pip install clawcrew

One-Click Install — Single curl command setup

Chain Command — Run multiple agents with auto context passing

Docker Support — Full containerized deployment

v0.2 Phase 3: Repo Mode

Repo Mode — Work with external GitHub repositories using Unified Diff patches

GitHub Integration — Analyze repos, manage issues, create PRs

Multi-Model Design — Design with multiple AI models, synthesize results

View full changelog →