LvcidPsyche/auto-browser

Give your AI agent a real browser — with a human in the loop. Open-source MCP-native browser agent.

What it solves

Auto Browser provides a secure, controlled way for AI agents to interact with the live web. It solves the problem of "brittle" web automation by allowing humans to step in and take over a live session when an agent gets stuck, while providing safety rails like approval gates, PII scrubbing, and audit trails to prevent unauthorized or dangerous actions.

How it works

It acts as an MCP-native (Model Context Protocol) control plane that manages Playwright-backed browser sessions. The system consists of a controller API that handles requests from LLM agents or humans and a browser node that runs Chromium. It supports multiple integration methods, including HTTP and a stdio bridge, allowing it to connect to clients like Claude Desktop, Cursor, and VS Code. Key features include noVNC for real-time visual takeover, named auth profiles to save and reuse login states, and a convergence harness for recording and verifying agent skills.

Who it’s for

  • Developers building MCP-powered agent workflows that require a real browser.
  • QA engineers performing operator-assisted browser debugging.
  • Teams managing internal dashboards or admin tools via AI agents.
  • Users who need to automate workflows on sites requiring persistent authentication.

Highlights

  • Human-in-the-loop: Live session takeover via noVNC when agents encounter issues.
  • Auth Persistence: Save and reuse named authentication profiles to avoid repeated logins.
  • MCP-Native: Built from the ground up for the Model Context Protocol, supporting various LLM clients.
  • Enterprise Safety: Includes Ed25519-signed Witness receipts, PII scrubbing, and strict compliance templates (e.g., strict, balanced).
  • Skill Induction: A convergence harness that records verified browser traces to create staged, signed skill candidates.

Related

  • Project
    anomalyco/browser-control

    Browser Control is an open‑source Node.js driver that lets AI agents or CLI users run Playwright code inside their own Chromium browser (with all logged‑in sessions and extensions). It works via a local relay and a browser extension, supports persistent sessions, inspection helpers, human hand‑offs, video/ HAR recording, and a TypeScript Effect client for secret‑safe API calls. Intended for trusted local use only.

  • Project
    vibheksoni/stealth-browser-mcp

    A stealth browser automation server for MCP-compatible AI agents that bypasses anti-bot protections like Cloudflare while providing deep browser control and network inspection.

  • Project
    remorses/playwriter

    A browser automation tool that lets AI agents control your existing, logged-in Chrome browser via an extension, enabling seamless interaction with authenticated websites.

  • Project
    AgentDeskAI/browser-tools-mcp

    BrowserTools MCP is an MCP server paired with a Chrome extension that lets AI coding agents observe the developer's actual browser session - console output, network activity, screenshots, and Lighthouse audits - without spinning up a separate automated browser. It keeps the agent in sync with the page you're already on, including your logins and state.

  • Project
    SawyerHood/dev-browser

    A browser automation tool that allows AI coding agents to maintain persistent browser sessions and use compact accessibility trees for efficient web interaction.