MCP server for browser automation via Vercel's agent-browser CLI
If you already use Agent, the agent mcp server is the piece that lets your assistant work with it directly. MCP server for browser automation via Vercel's agent-browser CLI.
Let LLMs drive a real browser. This MCP server wraps agent-browser (by Vercel Labs) so any MCP-compatible client — Claude Desktop, Claude Code, or others — can navigate pages, fill forms, click buttons, take screenshots, and more.
The toolset is worth reading before you wire it up, because it tells you what the integration is really for:
Navigation — open URLs, go back/forward, reloadInteraction — click, type, fill, press keys, hover, scroll, drag, select, check/uncheck, upload, downloadReading — get page text, HTML, attributes, URL, titleAccessibility — browser_snapshot returns an accessibility tree with @ref identifiers that can be used as selectors in subsequent callsState — check if elements are visible, enabled, or checkedCapture — screenshots (with optional annotation for vision models) and PDF exportSessions — create isolated browser sessions with independent cookies/storage/viewportCookies — get, set, clearJavaScript — evaluate arbitrary JS in the page contextDevTools — read console logs, inspect network requests, connect via CDPAGENT_BROWSER_PATH — Path to the agent-browser binaryConfiguration is passed through the environment: AGENT_BROWSER_PATH. Treat anything key-shaped as a real credential — scope it to the minimum the server needs, and rotate it if it ever lands in a shared config.
Install the agent-browser CLI:
Setup follows the usual MCP pattern — install or clone the server, register it in your client's configuration file, restart the client. The configuration blocks on this page cover the common clients.
This sits in the browser automation group, where several servers overlap in what they claim to do but differ sharply once you actually set them up. Agent's toolset — Navigation, Interaction, Reading and 8 more — is a fair guide to whether it matches your workflow. It is maintained by codeChap; worth a glance at recent repository activity before you build anything load-bearing on it.
This entry was verified against Agent's own documentation before publication; SyncDev keeps the directory reviewed rather than auto-generated.
| Tool | What it does |
|---|---|
| Navigation | open URLs, go back/forward, reload |
| Interaction | click, type, fill, press keys, hover, scroll, drag, select, check/uncheck, upload, download |
| Reading | get page text, HTML, attributes, URL, title |
| Accessibility | browser_snapshot returns an accessibility tree with @ref identifiers that can be used as selectors in subsequent calls |
| State | check if elements are visible, enabled, or checked |
| Capture | screenshots (with optional annotation for vision models) and PDF export |
| Sessions | create isolated browser sessions with independent cookies/storage/viewport |
| Cookies | get, set, clear |
| JavaScript | evaluate arbitrary JS in the page context |
| DevTools | read console logs, inspect network requests, connect via CDP |
| AGENT_BROWSER_PATH | Path to the agent-browser binary |
{
"mcpServers": {
"agent-browser": {
"command": "/path/to/mcp-server-agent-browser"
}
}
}Configuration as documented by the project. Restart the client after saving.
Install the agent-browser CLI:
| Variable | Description | Required |
|---|---|---|
| AGENT_BROWSER_PATH | Filesystem location the server is allowed to use. | Optional |
Microsoft's official browser automation server — drive a real browser through the accessibility tree, no screenshots needed.
Industrial-strength web extraction — render, scrape, crawl and search entire sites into clean markdown.
The original Chromium automation reference server — simple, screenshot-driven browser control.
Give your coding agent the full DevTools toolbox: traces, network, console, heap snapshots and Lighthouse.
Puppeteer-powered browser control that drives pages from the accessibility tree instead of pixels.
Cloud browsers for AI agents — automation sessions that run in Browserbase's fleet, not on your machine.