OpenClaw is a personal AI assistant that runs on your own computer or server. At its core it is an open source gateway built on Node.js that bridges the messaging apps you already use — Discord, Slack, Telegram, iMessage, WhatsApp — with AI agents. Instead of opening a dedicated chat window in a browser, you simply talk to your usual chat app and ask the AI to do the work. The project was started by Peter Steinberger in Austria and is now developed under the non-profit OpenClaw Foundation. It is MIT licensed, so the software itself is free to use.
Key Features
- Call AI from your chat apps: Supports many channels including Discord, Slack, Telegram, iMessage, WhatsApp, Signal, Microsoft Teams, Google Chat and Matrix. It can join both direct messages and group chats, and multiple channels can stay connected at the same time
- Self-hosted, so your data stays with you: It runs as a single process on your own machine rather than routing through a vendor’s servers. It works on macOS, Linux and Windows (PowerShell or WSL2), with several installation paths including Docker, npm and Nix
- Multi-agent routing and session management: You can assign different agents to different purposes and keep their sessions isolated, while preserving conversation context and per-user settings
- Tool use, file access and browser control: Beyond replying, agents can read and write files, run commands and browse the web. Plugins let you add further capabilities
- Web Control UI dashboard: A management screen opens locally at
http://127.0.0.1:18789/, where you can review chats and adjust settings from the browser. CLI commands such asopenclaw gateway statusandopenclaw triagecover status checks and diagnostics - Images, audio and documents: Media files are supported in addition to text, so you can send a file from your phone over chat and have it processed
Pricing
| Plan | Price | What it covers |
|---|---|---|
| OpenClaw itself | $0 (MIT license) | All features free to use; source code published on GitHub |
| Model usage | Per your AI provider’s pricing | A separate API key or subscription with Anthropic, OpenAI, Google, etc. is required |
Pricing is as of August 2026. Check the official site for the latest information.
OpenClaw has no paid tier of its own, but you need an API key for some AI model to run it. The setup wizard asks you to pick a provider and configure authentication. Your real costs are the usage-based or subscription fees of the model you choose, plus electricity or server costs if you keep it running around the clock.
Pros & Cons
✅ Pros
- Your everyday chat app becomes the entry point to AI, so there is no new app to learn
- Because it runs on your own machine, you control where your data lives
- MIT licensed open source, with no fee for the software itself
- Multiple channels and agents are managed through a single gateway
- Docker, npm, Nix and other installation options make it easy to fit into your environment
⚠️ Cons
- Being self-hosted, installation, keeping it running and updates are all on you
- Node.js version requirements (22.22.3+, 24.15+ or 25.9+, with 26 recommended) make the setup bar non-trivial
- Connecting each chat app (creating bots, linking accounts) follows a different procedure per channel
- Since agents can touch files and run commands, a careless permission design has a wide blast radius
- Development moves fast, so features and configuration details can change
Comparison with Similar Services
| Criteria | OpenClaw | Open WebUI | LibreChat | n8n |
|---|---|---|---|---|
| Main purpose | Bridge between chat apps and AI agents | Self-hosted AI chat UI | Self-hosted AI chat UI (multi-user) | Workflow automation |
| Entry point | Discord, Slack, Telegram, iMessage, etc. | Dedicated browser screen | Dedicated browser screen | Trigger nodes (webhooks, etc.) |
| Runtime | Your own machine (Node.js) | Self-hosted (Docker, etc.) | Self-hosted (Docker, etc.) | Self-hosted or cloud |
| Strength | Always-on personal assistant | Model switching and document reference | Permissions and team use | Automating many service integrations |
| Software cost | Free (MIT) | Free (OSS) | Free (OSS) | Free (self-hosted) / paid cloud |
Who Is It For
- People who want to ask an AI running on their home machine to do work from a phone chat app while out
- People who want their generative AI interactions to stay in an environment they control
- Teams or families who want a shared AI assistant sitting in a group chat
- People comfortable with the command line and server operations, who do not mind setting up their own environment
- Developers who want to use several AI agents for different purposes
Summary
OpenClaw is a gateway for moving “where you talk to AI” from a dedicated browser screen into the chat apps you already use. Being self-hosted, you decide where your data lives, and the software itself is free under the MIT license. In exchange, installation and operations are your responsibility, and it assumes you can handle Node.js setup and per-channel integration work. A realistic starting point is to connect just one easy channel such as Telegram or Discord, and expand toward an always-on setup once it feels like a fit.