ProductFeaturesSelf-Hosted AIHow it worksPricingDocsBlog
Sign inDownload
Self-Hosted AI / Personal AI agent

OpenClaw

A 24/7 AI agent that lives in WhatsApp, Telegram, or Discord - and can actually do things.

Download DeployOSSee all AI templates

What is OpenClaw?

OpenClaw is a self-hosted personal AI agent that connects to WhatsApp, Telegram, or Discord - the apps you already have open - rather than requiring you to visit a separate chat website. Message it like you'd message a person.

Beyond conversation, it can operate on your server: reading and writing files, running shell commands, and acting on requests you give it in plain language, so it functions closer to an assistant with hands than a Q&A chatbot.

Because it's self-hosted, the agent's memory, its command history, and the credentials it uses stay on your own infrastructure rather than a vendor's cloud.

How DeployOS deploys it

  1. 1DeployOS deploys the ghcr.io/openclaw/openclaw image with a persistent volume at /home/node/.openclaw for its state and configuration.
  2. 2A gateway token (OPENCLAW_GATEWAY_TOKEN) is generated automatically per install and stored in DeployOS's secret store - the same place your database passwords and API keys live - so the instance is authenticated from first boot without you generating or pasting a token by hand.
  3. 3Its gateway is published on port 18789 behind your reverse proxy's automatic HTTPS, then you connect it to WhatsApp/Telegram/Discord using that platform's usual bot-connection flow.
Imageghcr.io/openclaw/openclaw:latest
Port18789
Persistent volume/home/node/.openclaw
OPENCLAW_GATEWAY_TOKENauto-generated per install

Pairs well with

OllamaQdrantLetta (MemGPT)

Common questions

Is it safe to let an agent run shell commands on my server?

Treat OpenClaw's gateway token like a root credential - anyone with it can ask the agent to act on your server. DeployOS keeps it in the encrypted secret store rather than a plaintext env file, and never prints it to logs, but the operational judgment of what you authorize the agent to do is still yours.

Which messaging platforms does it support?

WhatsApp, Telegram, and Discord out of the box, per the shipped template - each connects through that platform's own bot/API setup once the OpenClaw instance is running.

Does it need its own LLM, or does it use Ollama?

It's commonly paired with a self-hosted model (Ollama) so the reasoning behind its responses stays on your infrastructure too - deploy both on the same server and they share the private Docker network.

Deploy OpenClaw on your own server in one click.

Free download, no account needed. AI templates are on paid plans.

Download DeployOS