MCP server for browser automation via Vercel's agent-browser CLI
MCPpedia last refreshed this data
io.github.codeChap/agent-browser is an MCP server that MCP server for browser automation via Vercel's agent-browser CLI. Its tool list has not been published yet over stdio and http, requires no API key, and scores 37/100 on MCPpedia's security, maintenance and efficiency rubric.
Config is the same across clients — only the file and path differ.
{
"mcpServers": {
"agent-browser": {
"command": "/path/to/mcp-server-agent-browser"
}
}
}Are you the author?
Add this badge to your README to show your security score and help users find safe servers.
Status: Testing — Functional but still being validated. Expect rough edges.
No automated test available for this server. Check the GitHub README for setup instructions.
Five weighted categories — click any category to see the underlying evidence.
No known CVEs.
No package registry to scan.
Be the first to review
Have you used this server?
Share your experience — it helps other developers decide.
Sign in to write a review.
Others in browser
Chrome DevTools for coding agents
Monitor browser logs directly from Cursor and other MCP compatible IDEs.
🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.
MCP server paired with a browser extension that enables AI agents to control the user's browser.
MCP Security Weekly
Get CVE alerts and security updates for io.github.codeChap/agent-browser and similar servers.
Start a conversation
Ask a question, share a tip, or report an issue.
Sign in to join the discussion.
Status: Testing — Functional but still being validated. Expect rough edges.
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.

Install the agent-browser CLI:
cargo install agent-browser
agent-browser install
cargo build --release
Binary: target/release/mcp-server-agent-browser
Add to your MCP client config (example for Claude Desktop / Claude Code):
{
"mcpServers": {
"agent-browser": {
"command": "/path/to/mcp-server-agent-browser"
}
}
}
The server communicates over stdio using JSON-RPC 2.0 — it works with any MCP client.
| Environment Variable | Description | Default |
|---|---|---|
AGENT_BROWSER_PATH | Path to the agent-browser binary | agent-browser (via $PATH) |
The server exposes 35+ tools covering:
browser_snapshot returns an accessibility tree with @ref identifiers that can be used as selectors in subsequent callsEach tool is self-documented with parameter schemas — your MCP client will discover them automatically.
A typical AI-driven interaction follows this pattern:
browser_navigate — open a pagebrowser_snapshot — get the accessibility tree to understand the page structurebrowser_click / browser_fill — interact using @ref selectors from the snapshotbrowser_screenshot — visually verify the resultThe agent decides which tools to call and in what order. You just describe what you want done.
All tools accept an optional session_id. Create isolated sessions with browser_new_session and pass the returned ID to subsequent calls. Each session has its own cookies, storage, and viewport. Without a session ID, commands use the default session.
Commands time out after 60 seconds by default. The browser_wait tool adjusts its timeout automatically based on the wait duration you specify.
agent-browser daemonagent-browser uses Playwright under the hoodMIT