OpenClaw is an open-source, MIT-licensed project that sits between a messaging platform — Discord, Slack, Telegram — and an LLM agent loop. Instead of building your own bot from scratch to read messages, call a model, and post a reply, OpenClaw already does that wiring: you point it at a model and it becomes a "digital coworker" living in your server, able to read channels, hold context across a conversation, and (depending on what you connect it to) take actions like running a tool or posting a scheduled update.
| OpenClaw provides | You still need |
|---|---|
| The bridge: platform events → agent loop → reply | An LLM API key (or a self-hosted model) to act as the brain |
| Conversation memory / context handling | A server to run it on (or a laptop, for testing) |
| A plugin/tool system for actions beyond chat | Your own judgement on what the bot is allowed to touch |
The software is free because it's genuinely open source — there's no OpenClaw subscription, no license fee, no usage cap imposed by the project itself. That's exactly why lesson 2 matters: "free" describes the gateway, not the electricity bill of the model running behind it.
Knowledge check
OpenClaw is free and open-source. Does that mean running an OpenClaw bot costs nothing?