Webcmd is browser infrastructure designed for AI agents that learns from real browser interactions and compiles that knowledge into reusable, structured commands. It reduces token consumption by up to 90% by replacing repeated screenshot and DOM analysis with deterministic skills, sitemaps, and APIs. The system progresses from live browser control through memory layers until many tasks no longer require a full browser session. It is distributed as an npm package and targets developers building reliable, cost-efficient agentic web automation.
In the Developer Tools space, Webcmd takes a focused approach. AI agents repeatedly rediscovering how to interact with websites, leading to high token usage, fragility, and inconsistent performance. It is built as an open-source project for developers. The project is open source (Apache-2.0). Webcmd is available on the command line, and it can be self-hosted.
Behind Webcmd is agentrhq, and it first shipped in 2026. Development happens publicly on GitHub with 102 stars and 369 commits in the last 90 days. Among its 6 catalogued features are Browser Control, Skill Learning, and Structured Output.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do