N8n Claw

n8n-claw is an open-source, self-hosted AI agent built entirely on n8n workflows. Inspired by OpenClaw, it provides an autonomous assistant that communicates via Telegram, Slack, Teams, or a webhook API. The agent features adaptive RAG-powered memory with hybrid search (semantic, full-text, entity), a knowledge graph, and nightly pattern extraction to model user behavior. It can install pre-built MCP skills from a catalog of 70+ integrations, manage tasks and reminders, delegate to specialized sub-agents (Expert Agents), and perform web research with Chromium browser automation including 2FA via Telegram round-trip. Built for developers and teams who want full control over their AI infrastructure, n8n-claw runs entirely on your own server with a one-script setup. It supports eight LLM providers out of the box (Claude, GPT, Gemini, Mistral, DeepSeek, Ollama, OpenRouter, or any OpenAI-compatible endpoint). The agent can connect Google services via OAuth2, send voice messages, analyze photos, extract documents, and even delegate coding tasks to an OpenClaw instance. Key differentiators include proactive task management with smart reminders and recurring actions, project memory via persistent markdown documents, and smart background checks that monitor silently. The skill catalog allows natural-language installation of integrations like HubSpot, Salesforce, Jira, Stripe, and more. No cloud lock-in and no manual config files — all setup is handled through a single script and Telegram chat.