OpenClawd

Multi-provider AI platform. 20+ LLMs. MCP integrations. Desktop + messaging. Voice input. Token analytics.

Architecture

npx openclawd-cli

Install

git clone https://github.com/rohitg00/openclawd
cd openclawd && ./setup.sh

Add your API key to .env:

ANTHROPIC_API_KEY=sk-ant-xxx

Run:

cd server && npm start   # Terminal 1
npm start                # Terminal 2

Providers

Works with any provider. Add API keys to .env:

ANTHROPIC_API_KEY      # Claude (default)
OPENAI_API_KEY         # GPT-4o, o1
GEMINI_API_KEY         # Gemini 2.0
GROQ_API_KEY           # Ultra-fast Llama
OPENROUTER_API_KEY     # 200+ models
VENICE_API_KEY         # Privacy-focused
DEEPSEEK_API_KEY       # DeepSeek R1
XAI_API_KEY            # Grok
MISTRAL_API_KEY        # Mistral/Codestral

Ollama models auto-discovered when running locally.

Model Fallback

Automatic failover when a provider fails. Configure in Settings > General or .env:

FALLBACK_PROVIDER=openai
FALLBACK_MODEL=gpt-4o

When the primary provider errors out, OpenClawd automatically retries with the fallback.

Token Usage Dashboard

Settings > Usage tab shows:

• Summary cards (total tokens, estimated cost, request count)
• 7-day token usage bar chart
• Per-provider breakdown table
• Per-message token badge in chat

Channel Bots

Connect AI to messaging platforms. Configure in Settings > Channels or .env:

TELEGRAM_BOT_TOKEN=xxx        # Telegram (polling mode)
DISCORD_BOT_TOKEN=xxx         # Discord (intents)
SLACK_BOT_TOKEN=xoxb-xxx      # Slack (Socket Mode)
SLACK_APP_TOKEN=xapp-xxx      # Slack app token

WhatsApp connects via QR code scan (Baileys, no token needed).

Each bot supports /reset, /model <name>, and /status commands. Per-user conversation history with 50-message cap and 24h session expiry.

Voice Input

Click the mic button or press Ctrl/Cmd+Shift+V to dictate. Uses the browser Web Speech API — no external services needed. Supports continuous dictation with interim results.

MCP Servers

20+ servers included. Add via API or config:

# Add from catalog
curl -X POST localhost:3001/api/mcp/servers/from-catalog/github

# Or edit server/mcp-servers.json

Available: filesystem, memory, git, github, slack, postgres, puppeteer, and more.

API

# Chat
curl -X POST localhost:3001/api/chat \
  -H "Content-Type: application/json" \
  -d '{"message": "Hello", "chatId": "test"}'

# Providers
curl localhost:3001/api/llm/providers

# MCP catalog
curl localhost:3001/api/mcp/catalog

# Usage history
curl localhost:3001/api/llm/usage/history

# Channel status
curl localhost:3001/api/channels/status

# Start a channel bot
curl -X POST localhost:3001/api/channels/telegram/start \
  -H "Content-Type: application/json" \
  -d '{"token": "your-bot-token"}'

Detailed Architecture

User Interfaces

• Desktop App — Electron-based native app for macOS, Windows, Linux with voice input and token dashboard
• CLI — npx openclawd for terminal-first workflows
• Channel Bots — Telegram, Discord, WhatsApp (QR), Slack (Socket Mode)

Backend (Server :3001)

• Chat API — SSE-streamed responses with model fallback
• Provider Registry — Unified interface to 20+ LLM providers with automatic failover
• Channel Manager — Bot lifecycle, session management, per-user conversation state
• MCP Manager — Loader + catalog for 30+ MCP server integrations
• Auth Profiles & Usage — API key management, token tracking, cost estimation
• Agent Loop & Memory — Persistent conversation context and agent orchestration

LLM Providers (20+)

Claude (Agent SDK), OpenCode (Free+Paid), Ollama (Local), OpenAI, Gemini, Groq, DeepSeek, Mistral, xAI, and more.

MCP Servers (30+ catalog)

Filesystem, Memory, GitHub, Fetch, Brave, Slack, Git, DB, and more. Add from the built-in catalog or configure manually.

Build & Release (GitHub Actions)

• macOS .dmg / .zip
• Windows .exe (NSIS)
• Linux .AppImage / .deb
• npm openclawd-cli
• GitHub Release (Draft)

License

MIT