n8n-claw
by freddy-schuetz·★ 436·Score 51
Self-hosted AI agent using n8n with MCP Skills library, memory management, and expert agents.
Overview
n8n-claw is a comprehensive self-hosted AI agent built on n8n, PostgreSQL, and Claude. It features adaptive RAG-powered memory, a library of pre-built MCP skills, and expert agents with delegated sub-agents. The agent can be accessed via Telegram or HTTP API, offering proactive task management, media understanding, and browser automation capabilities. It's designed as a complete solution that runs on your own infrastructure with a single setup script.
Try asking AI
After installing, here are 6 things you can ask your AI assistant:
When to choose this
Choose this if you want a complete AI agent solution with task management and memory capabilities that integrates MCP skills as one feature.
When NOT to choose this
Don't choose this if you specifically need an MCP server to expose tools/resources to other systems; this is a consumer agent, not a provider of MCP services.
Tools this server exposes
12 tools extracted from the READMEquery_postgresQuery PostgreSQL database directly
web_searchSearch the web via self-hosted SearXNG instance
web_readerRead webpages as clean markdown
memory_searchSearch long-term memory using semantic or keyword search
add_taskCreate a new task with priority and due date
set_reminderSet a timed reminder for a specific time
send_telegram_messageSend a message via Telegram
browse_websiteAutomate browser actions using Browser Use
file_downloadDownload files from the internet or cloud services
schedule_actionSchedule an action to be executed at a specific time
query_knowledge_graphQuery the knowledge graph for entity relationships
build_mcp_skillBuild custom MCP skills from scratch
Comparable tools
Installation
Installation
Prerequisites
- Linux VPS (Ubuntu 22.04/24.04 recommended, 4GB RAM and 15GB Disk minimum)
- Telegram Bot Token (from [@BotFather](https://t.me/BotFather))
- Your Telegram Chat ID (from [@userinfobot](https://t.me/userinfobot))
- LLM API Key (Anthropic, OpenAI, OpenRouter, DeepSeek, Gemini, Mistral, Ollama, or OpenAI-compatible)
- Domain name (or use [sslip.io](https://sslip.io))
Setup Steps
git clone https://github.com/freddy-schuetz/n8n-claw.git && cd n8n-claw && ./setup.shThe script will prompt for:
- n8n API Key (generated in n8n UI: Settings → API)
- Telegram Bot Token and Chat ID
- LLM API Key (choose your provider)
- Domain name
- Agent personality (name, language, communication style)
Optional features:
- Embeddings for semantic memory search
- Voice messages (requires Whisper API key)
Claude Desktop Integration
{
"mcpServers": {
"n8n-claw": {
"command": "docker",
"args": ["run", "--rm", "-i", "n8n-claw-agent", "mcp"],
"env": {
"TELEGRAM_BOT_TOKEN": "your-telegram-token",
"TELEGRAM_CHAT_ID": "your-chat-id",
"LLM_API_KEY": "your-llm-key",
"DOMAIN": "your-domain"
}
}
}
}After setup, send a message to your Telegram bot to start interacting.
FAQ
- What LLM providers does n8n-claw support?
- n8n-claw supports Anthropic (default), OpenAI, OpenRouter, DeepSeek, Google Gemini, Mistral, Ollama (local), and any OpenAI-compatible endpoint.
- How secure is n8n-claw?
- The agent can be secured by restricting Telegram access to your chat ID only via the n8n UI. All communication uses HTTPS webhooks. Credentials are stored in environment variables with restricted access.
- Can I integrate n8n-claw with other tools?
- Yes, it provides a webhook API that can be integrated with Slack, Teams, or custom applications. It also has an MCP Skills library for extending capabilities.
Compare n8n-claw with
Last updated · Auto-generated from public README + GitHub signals.