What is OpenClaw?
OpenClaw is a self-hosted Gateway for AI agents. It connects chat apps and channel surfaces to agent workflows, so you can send a message from the tools you already use and receive an agent response back through the same surface.
OpenClaw in one sentence
OpenClaw gives AI agents a self-hosted communication layer: the Gateway listens across configured channels, keeps track of sessions and routing, and lets the agent operate from familiar chat surfaces instead of a single closed interface.
Gateway first
OpenClaw centers on a Gateway: the long-running service that owns channel connections, sessions, routing, nodes, and hooks.
Chat apps as control surfaces
OpenClaw connects agent workflows to chat surfaces such as Discord, Slack, Telegram, WhatsApp, Signal, iMessage, Google Chat, Matrix, WebChat, and more.
AI agents behind the conversation
The Gateway routes messages to AI coding agents and returns responses where you already communicate, instead of forcing every workflow into a single dashboard.
Operational control
OpenClaw is designed for builders who care about configuration, sessions, workspace files, provider keys, and the runtime that keeps the agent online.
How OpenClaw works
The official docs describe the Gateway as the center of OpenClaw. Hosting decisions mostly come down to where that Gateway runs and who operates it.
Run the OpenClaw Gateway
The Gateway is the service layer. In a self-hosted setup, you install and operate it on your own machine, server, or VPS.
Connect channels
Each chat channel connects through the Gateway. You can configure one or multiple channels depending on where you want to talk to the agent.
Bring provider access
OpenClaw workflows need model provider access. BYOK setups let teams keep direct control over keys, billing, and model selection.
Keep the agent reachable
For production use, the Gateway needs uptime, logs, updates, safe remote access, and enough isolation for private workspace activity.
When self-hosting fits
You want complete control over the host, network, package versions, and security posture.
You are comfortable operating a Node-based service, channel credentials, logs, upgrades, and incident recovery.
Your usage is experimental or local-first, so temporary downtime is acceptable.
When managed hosting fits
You want OpenClaw online without becoming the server operator.
You need a private hosted workspace for long-running agent workflows, terminal access, files, and channel availability.
You prefer a managed path for infrastructure while still keeping control of provider keys and workspace behavior.
OpenClaw vs. GetClaw
Frequently asked questions
Is OpenClaw an AI model?
No. OpenClaw is a self-hosted gateway and runtime layer for connecting chat apps and channel surfaces to AI agents. You still choose the model/provider access used by your agent workflow.
Is OpenClaw only for one chat app?
No. The official docs describe support for many channel surfaces, including Discord, Slack, Telegram, WhatsApp, Signal, iMessage, Google Chat, Matrix, WebChat, and plugin-based channels.
Do I need hosting for OpenClaw?
Not always. You can self-host OpenClaw. Hosting becomes useful when you want the Gateway available beyond your laptop, especially for private agent workflows that need uptime and remote channel access.
What does GetClaw add?
GetClaw provides managed OpenClaw hosting: a private hosted workspace for running OpenClaw-style agent workflows with infrastructure, terminal access, files, BYOK support, and deployment-oriented defaults.
Want to host OpenClaw instead of operating it manually?
Compare GetClaw plans or read the dedicated OpenClaw hosting page to choose the right path for your private agent workspace.