Loading your favourites…
PUBLIC FULL SUMMARY · Repository
OpenClaw
openclaw/openclaw
AI editorial summary
OpenClaw is an open-source AI assistant that runs on your own computer and works through chats such as WhatsApp, Telegram, Slack and Discord. It can research topics, edit files, operate a browser and schedule work using enabled tools. Its background service connects chats, AI models and tools. State and credentials stay on your hardware, while configured external services receive the requests needed for their work. documented capabilities; available tools depend on configuration · Assistant · Tool categories. Project guide
What it does
Illustrative use case: ask through Telegram for a short brief on a topic you need to understand. With web tools enabled, the assistant could find sources and return a brief with links in the chat for you to review. Individuals or configured teams can also give it work through its browser interface, called the Control UI. This scenario combines documented tools and channel support. documented components; combined research scenario is editorial inference · Assistant · Tool categories.
Engineering relevance
A background service called the Gateway connects conversations to the AI model and available tools. Tool policy, provider restrictions, sandbox settings and channel permissions determine which actions the assistant can request. Potential engineering use: design a research assistant that can search sources while withholding file changes or messages to other channels. The tool overview describes these controls; a specific deployment needs the detailed policy and security documentation. documented mechanisms; proposed access policy is editorial inference · Connections · Tool policy.
How to set it up
See the Setup Guide. The captured guide requires Node.js 24.16+ or 26.1+ and an existing supported coding-agent login or provider API key. Windows options include the Hub app, PowerShell installer and WSL2; macOS and Linux paths are also documented. Quick start can detect existing AI access and makes a real AI completion during onboarding, so setup itself can consume provider or subscription usage. Main-session tools run on the host unless sandboxing is configured. Review tool access and service charges before connecting channels. documented setup and usage implications · Requirements · Onboarding · Host access.
Next investigation
A possible first evaluation is a brief on one topic through the browser interface. After agreeing any setup and AI usage, check whether the links support the brief and whether the assistant stays within the enabled research tools. Review the detailed tool policy and sandbox configuration before permitting file changes or messages to others. editorial inference · Tool policy · Onboarding · Host access. Developer references.