Gives Cursor's agent a real browser via the Browser Use CLI. Bundles:
- Skill (
skills/browser-use/SKILL.md): teaches the agent to drive Chrome (or a Browser Use Cloud browser) through thebrowser-useCLI — navigation, coordinate clicks, extraction, screenshots, cloud daemons. - MCP server (
mcp.json): registersuvx browser-use@latest --cli-mcp, a self-installing stdio server exposingbrowser_execandbrowser_screenshotwith a persistent Python helper namespace.
Only prerequisite on the user's machine: uv.
From the Cursor Marketplace: search for Browser Use and install.
For local development, symlink this directory into Cursor's local plugins folder and reload the window:
ln -s /path/to/plugins/cursor ~/.cursor/plugins/local/browser-useTeams/Enterprise orgs can also import this repo as a team marketplace (Dashboard → Plugins → Team Marketplaces → Add Marketplace → Import from Repo); the root .cursor-plugin/marketplace.json points at this directory.
cursor/
├── .cursor-plugin/plugin.json # manifest
├── mcp.json # bundled MCP server config (stdio, self-installing via uvx)
├── skills/browser-use/SKILL.md # the skill
└── assets/logo.png
- Local Chrome control requires enabling remote debugging once at
chrome://inspect/#remote-debugging. - Cloud browsers need
browser-use auth login(the skill covers this).