Privacy-first
End-to-end on your hardware. Encrypted memory store, no third-party analytics, full audit log of every tool call.
OpenClaw is a self-hosted personal AI assistant that connects to all your channels — Telegram, Discord, WhatsApp and more. Powered by Claude, GPT, or any LLM. 100% yours.
Your conversations, files and memory never leave your server. No cloud middleman, no logs sold to anyone.
Talk to the same assistant from Telegram, Discord, the Web UI — context follows you across channels.
Single Docker command. Bring your own keys for Claude, OpenAI, or any local model. No vendor lock-in.
An assistant that lives where your data lives. Compose skills, swap models, and integrate any channel — without rewriting the stack.
End-to-end on your hardware. Encrypted memory store, no third-party analytics, full audit log of every tool call.
Native adapters for Telegram, Discord, WhatsApp, Signal, Slack, SMS — plus a sleek browser Control UI for power use.
Drop a Python or TypeScript file into skills/ and your claw can code, fetch weather, run healthchecks — anything.
Claude for reasoning. GPT for vision. A local Llama for cheap chat. Route per-skill, per-channel, or per-message.
One docker compose up deploys the runtime, Control UI, queue and vector store. Works on a $5 VPS.
MIT licensed. Fork it, audit it, sell it, ship your own. The community ships a new skill every week.
From a fresh VPS to a personal assistant pinging you on Telegram — no DevOps PhD required.
Pull the container, set your domain, hit go. Caddy handles HTTPS automatically.
$ curl -fsSL claw.sh | sh
Paste your Telegram or Discord token in the Control UI. Adapters connect in seconds.
$ claw adapter add telegram ↳ paste BOT_TOKEN
Send a message from any channel. Add skills as you grow — it's just files in a folder.
@you remind me to water the plants at 7 ✓ scheduled
One command, one $5 server, infinite possibilities. The assistant that answers to you — and only you.