Webcmd
A local browser automation tool for AI agents that remembers websites and turns repeated workflows into reusable commands, licensed under Apache 2.0.
Webcmd is a browser automation tool for developers whose AI agents repeatedly work with the same websites. It keeps what agents learn locally so later tasks can use familiar workflows instead of inspecting each site from scratch.
- Live browser control: Agents can inspect pages, click controls, enter text, extract information and capture network calls in a real browser.
- Website memory: Webcmd retains observed pages, actions, workflows, APIs and fallback paths for reuse during later tasks.
- Reusable commands: Site adapters turn recurring workflows into commands with named inputs and structured results; some can complete tasks without opening a browser.
- Authenticated browsing: Agents can work through logged-in profiles on sites such as LinkedIn, PubMed, Hacker News and ChatGPT.
- Agent-readable results: Commands return machine-readable output and errors, while a registry exposes their inputs and browser requirements.
Webcmd runs locally, requires Node.js, and works with Claude, Codex and other supported agent harnesses. Website tasks access the sites involved; local operation doesn't make those services offline. First access may use a Webcmd Cloud seed, while subsequent learning stays local. The live browser takes precedence over stored memory, and a memory failure doesn't block the task. The project is open source under Apache 2.0.
