mula-browser

# mula-browser — AI-first anti-detect browser with token-efficient extraction > An AI-first browser combining anti-detect fingerprinting with human behavior simulation. Drive it via CLI, REST API, or MCP server — and extract pages in ~500–800 tokens. [](https://github.com/NachaFromMars) ## Overview mula-browser (mubro) is an AI-first browser built on Electron with a Hono ACP server on port 9870. It pairs anti-detect fingerprinting with human behavior simulation to browse naturally, while keeping page extraction token-efficient (~500–800 tokens/page). Control it through a CLI script, a local REST API, or an MCP stdio server for Claude/Cursor integration. ## Features - **Anti-detect fingerprinting** + human behavior simulation - **Token-efficient extraction** — ~500–800 tokens/page - **REST API** at `http://127.0.0.1:9870/api` — status, navigate, back, forward, refresh, extract, click, screenshot - **MCP server** — `node dist/server/mcp.js` (stdio mode) ## Usage / Quick Start ```bash node scripts/mubro.mjs start # launch browser + ACP server node scripts/mubro.mjs nav <url> # navigate node scripts/mubro.mjs extract # extract page (~500-800 tokens) node scripts/mubro.mjs snap # screenshot node scripts/mubro.mjs status # health check # MCP mode (Claude/Cursor) node dist/server/mcp.js ``` ## Trigger Keywords (OpenClaw) mubro, mula browser, AI browser, anti-detect browser, human browser, browser extraction ## Related Skills - [crawl-for-ai](https://github.com/NachaFromMars/crawl-for-ai) — Crawl4AI JS-rendering scraper --- Part of the [NachaFromMars](https://github.com/NachaFromMars) OpenClaw skill ecosystem.