agent-browser is an open-source browser automation harness and agent tool developed by Vercel Labs. It provides autonomous AI agents with structured interfaces to navigate web pages, inspect DOM elements, execute actions, and parse visual or text content in headless and interactive browser environments. In day-to-day operations, agent-browser functions as a CLI utility, orchestration layer, and tool integration (including Claude plugin and skill integrations). It allows developers and autonomous coding agents to execute programmatic browser sessions, run end-to-end web evaluations, and verify web application states against LLM prompts. What distinguishes agent-browser from standard headless automation drivers like raw Playwright or Puppeteer is its dedicated agent abstraction layer. It includes built-in agent skill specifications, standardized evaluation harnesses (`evals`), and benchmarks tailored for testing how effectively LLMs can navigate complex web interactions. The project is available as an open-source repository on GitHub under permissive licensing, allowing teams to self-host and run it locally, in Docker environments, or within CI/CD pipelines.
Tags: ai agent, cli, developer tools, OpenSource, terminal