Browsentic: open-source MCP server lets AI coding agents drive your logged-in browser
NoEgg2932 · reddit · 2026-09-09
A developer open-sourced Browsentic (MIT), a browser extension plus local daemon that lets MCP clients drive the browser you're already logged into — beyond what Playwright or Claude in Chrome offer.
Architecture
- MV3 service workers can't listen for connections, so the extension dials out to a daemon on 127.0.0.1 that owns the browser link; MCP clients speak stdio to the same daemon.
- One command registers it: claude mcp add browsentic -- browsentic mcp. Claude Code, Codex, Cursor and Zed can all share one browser and interleave; tool calls stay correlated but shared page state needs careful design.
Security
- Two gates: peers classified by handshake Origin (which pages can't forge), then proven via pairing code or origin-bound session key with mutual nonce answers to prevent port squatting.
- MCP clients have no approval channel, so form submissions, uploads, captcha answers default to deny for external callers.
- Page text is fenced with per-daemon random markers and labeled as data, not instructions — narrowing the prompt injection surface, though not a fix.
More from coding & agent
- Claude Code Drives 46.5% of Coding-Agent Traffic to Hugging Face, Codex Second — vanstriendaniel · 2026-09-09
- AI Engineering Buildcamp Reveals Six Hands-On Projects, from RAG to Multi-Agent Systems — Al_Grigor · 2026-09-09
- What's the most common way people actually set up multi-agent coding workflows? — Necessary-Apple337 · 2026-09-09
- New CLI tool scrapes WeChat articles into Markdown from any post URL — vista8 · 2026-09-09
- The expensive part of coding agents isn't the agents—it's the silent retries ($900 for one task) — mrtrly · 2026-09-09
- A green test suite proves an agent can write tests, not that the feature works — Future_AGI · 2026-09-09