This project provides an MCP-native browser control plane for AI agents and operators, enabling sophisticated web automation with human oversight. It addresses the need for LLM agents to interact with real browsers, offering features like reusable authentication, human takeover, and robust safety rails, making it ideal for internal dashboards, assisted QA, and complex login workflows.

How It Works

Auto Browser leverages Playwright for browser automation, exposing its capabilities as an MCP server. This MCP-native design allows seamless integration with AI agents and other clients via HTTP or stdio. A key feature is the integration of noVNC, enabling human operators to take over live browser sessions when web interactions become brittle or require manual intervention. The system also supports saving and reusing named authentication profiles, streamlining workflows by eliminating repetitive logins.

Quick Start & Requirements

The primary method for local development is via Docker Compose:

git clone https://github.com/LvcidPsyche/auto-browser.git
cd auto-browser
docker compose up --build

This requires Docker and Docker Compose. For local readiness checks, run make doctor. Key endpoints include API docs at http://127.0.0.1:8000/docs, the operator dashboard at http://127.0.0.1:8000/dashboard, and visual takeover at http://127.0.0.1:6080/vnc.html. A hosted demo is available via Codespaces.

Highlighted Details

• Playwright-backed browser sessions offering screenshots, DOM summaries, OCR, tab control, downloads, and network inspection.
• Comprehensive operator safety features, including approval gates, operator identity verification, PII scrubbing, and audit trails.
• Flexible deployment options: Docker Compose for local setup, Codespaces for hosted demos, and MCP over HTTP/stdio bridge for integration.
• "Login once, reuse later" functionality via named authentication profiles.
• Seamless human takeover capability through noVNC for complex or unexpected web scenarios.

Maintenance & Community

The repository includes contribution guidelines (CONTRIBUTING.md) and code of conduct (CODE_OF_CONDUCT.md), suggesting an active development process. Specific community channels (e.g., Discord, Slack) or notable contributors/sponsors are not detailed in the README.

Licensing & Compatibility

The README does not explicitly state the software license. This omission requires further investigation for compatibility, especially for commercial use or integration into closed-source projects. The system is designed for integration with MCP clients and supports deployment via Docker and Codespaces.

Limitations & Caveats

This project is explicitly not intended for stealth operations, anti-bot measures, CAPTCHA solving, unauthorized scraping, or deceptive identity manipulation. Release v1.0.2 indicates ongoing development, and users should consult the CHANGELOG.md for specific release details and potential breaking changes.