FastClaw is a high-performance AI agent framework with multi-agent support, multi-channel integration, and comprehensive extensible features.

• Multi-Agent Orchestration: Coordinate multiple specialized agents with different roles
• Multi-Channel Support: 8+ channel adapters (Telegram, Slack, Discord, Feishu, DingTalk, QQ, WeCom, WhatsApp)
• Real LLM Integration: Support for OpenAI, Anthropic, Ollama, and DeepSeek
• Memory System: JSONL session storage + Markdown long-term memory + SQLite vector search
• Tools & Skills: Dynamic tool registration with sandbox execution and extensible skill system
• MCP Protocol: Model Context Protocol support for extensible tool integration
• Plugin System: Modular plugin architecture for custom extensions
• Audit Logging: Comprehensive audit trail for all operations
• Error Handling: Advanced error handling with retries and graceful degradation
• CLI Interface: Command-line interface for management and testing
• Workspace Management: Multi-workspace support for isolated environments

FastClaw implements a comprehensive agent framework:

• Gateway: WebSocket control plane for real-time communication
• Agent Runtime: Embedded agent loop with inference-execution cycle
• Multi-Agent Orchestrator: Coordinate multiple specialized agents
• LLM Integration: Multiple LLM providers with unified interface
• Memory System: JSONL session storage + Markdown long-term memory + SQLite vector index
• Tools & Skills: Dynamic tool registration with sandbox execution
• MCP Support: Model Context Protocol for extensible tool integration
• Plugin System: Modular plugin architecture
• Audit Logger: Comprehensive operation logging

fastclaw/
├── core/                 # Core framework
│   ├── gateway.py       # WebSocket gateway
│   ├── agent.py         # Agent runtime & loop
│   ├── llm_client.py    # LLM client integration
│   ├── tools.py         # Tool registry
│   ├── multi_agent.py   # Multi-agent orchestrator
│   ├── workspace.py     # Workspace management
│   ├── audit.py         # Audit logging
│   ├── error_handler.py # Error handling
│   ├── plugin.py        # Plugin system
│   ├── cli.py           # CLI interface
│   ├── llm/             # LLM adapters
│   │   ├── base.py
│   │   ├── openai_adapter.py
│   │   ├── anthropic_adapter.py
│   │   └── ollama_adapter.py
│   ├── mcp/             # MCP protocol
│   │   ├── manager.py
│   │   └── protocol.py
│   └── skills/          # Skills system
│       ├── base.py
│       ├── manager.py
│       └── registry.py
├── adapters/            # Channel adapters
│   ├── base.py
│   ├── telegram.py
│   ├── slack.py
│   ├── discord.py
│   ├── feishu.py
│   ├── dingtalk.py
│   ├── qq.py
│   ├── wecom.py
│   ├── whatsapp.py
│   └── webhook.py
├── storage/             # Storage implementations
│   ├── session_store.py
│   └── memory_store.py
├── workspace/           # Agent workspace (markdown files)
│   ├── AGENTS.md
│   ├── SOUL.md
│   ├── TOOLS.md
│   └── MEMORY.md
├── state/               # Runtime state
│   ├── sessions/
│   └── memory/
├── ui/                  # Web interface
│   └── index.html
├── main.py              # Entry point
├── config.yaml          # Configuration file
├── ARCHITECTURE.md      # Architecture documentation
├── ROADMAP.md           # Project roadmap
└── requirements.txt     # Python dependencies

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Set environment variables
export DEEPSEEK_API_KEY="your_api_key_here"
export DEEPSEEK_BASE_URL="https://api.deepseek.com/v1"

# Start the server
python3 main.py

# Access the web interface
# Open http://localhost:8000 in your browser

• POST /api/v1/chat - OpenAI-compatible HTTP API
• WS /api/ws - WebSocket endpoint for real-time communication
• GET /api/v1/tools - List available tools
• GET /api/v1/sessions - List active sessions
• GET / - Web interface

Edit config.yaml to configure:

llm:
  provider: "openai"  # openai | anthropic | ollama
  openai:
    api_key: "${DEEPSEEK_API_KEY}"
    base_url: "${DEEPSEEK_BASE_URL}"
    model: "deepseek-chat"

agents:
  - id: "coordinator"
    role: "coordinator"
    model: "deepseek-chat"
    enabled: true

channels:
  telegram:
    enabled: false
    bot_token: "${TELEGRAM_BOT_TOKEN}"
  slack:
    enabled: false
    bot_token: "${SLACK_BOT_TOKEN}"

• OpenAI: GPT-4, GPT-3.5
• DeepSeek: DeepSeek-V3, DeepSeek-Coder
• Anthropic: Claude 3 Opus, Sonnet, Haiku
• Ollama: Local models (Llama2, Mistral, etc.)

• Telegram: Full bot API support
• Slack: App integration
• Discord: Bot with slash commands
• Feishu: Enterprise bot
• DingTalk: Enterprise bot
• QQ: OneBot protocol
• WeCom: Enterprise bot
• WhatsApp: Business API
• Webhook: Generic webhook adapter

• bash: Execute shell commands (with approval)
• file_read: Read file contents
• file_write: Write file contents
• file_list: List directory contents
• calculator: Evaluate mathematical expressions
• memory_add: Add to long-term memory
• memory_search: Search memory

# List all CLI commands
python3 -m core.cli --help

# Start agent with specific configuration
python3 -m core.cli start --config config.yaml

# List available tools
python3 -m core.cli tools list

# Run a test query
python3 -m core.cli query "Hello, FastClaw!"

• Features Guide - Detailed configuration and usage instructions for all features
• Channel Integration Guide - Complete tutorials for 8+ communication channels
• API Reference - Complete reference for REST API and WebSocket API
• CLI Guide - Complete manual for command-line tools

• ARCHITECTURE.md - Detailed architecture documentation
• ROADMAP.md - Project roadmap and milestones
• DEEPSEEK_SETUP.md - DeepSeek integration guide

• README_ZH.md - 简体中文版 README | Simplified Chinese README

MIT License