DrissionPage MCP Server provides a browser automation backend for multimodal AI clients like Codex and Claude Code. It bridges AI vision understanding with real browser interactions by translating identified coordinates into natural pointer action chains, enhancing automation capabilities beyond traditional selector-based methods. This is beneficial for AI-driven UI automation, testing, and research.

How It Works

The server offers a dual approach: robust structured DOM automation via 52 tools and a novel vision-guided interaction layer. This layer converts multimodal AI's viewport coordinate outputs into physically plausible pointer action chains. It employs natural pointer dynamics, including cubic Bézier curves, human-like timing, and realistic click semantics, enabling interaction with visually identified elements where selectors are insufficient.

Quick Start & Requirements

• Install: python -m pip install -U drissionpage-mcp
• Prerequisites: Python 3.10+ (3.11+ recommended), Chrome or Chromium browser, and an MCP-compatible client (e.g., Codex CLI/IDE, Claude Code).
• Setup Time: Approximately 30 seconds for Codex CLI/IDE integration.
• Docs: README.md, Compatibility, Tool Contract, Troubleshooting.

Highlighted Details

• Vision-Guided Interaction: Translates AI vision coordinates into natural pointer movements (Bézier curves, human-like timing, physical click semantics) for precise UI control.
• Structured Automation: Provides 52 tools for deterministic element selection using CSS/XPath normalization and interaction.
• Quality & Validation: Backed by strict regression suites, browser-integrated scenario checks, and a 95% code coverage floor enforced in CI.
• Type-Safe: Utilizes Pydantic models and full type hints for all tools, ensuring robust data handling.

Maintenance & Community

The project is actively maintained by Wukunyun. While specific community channels like Discord/Slack are not listed, contribution guidelines are available in CONTRIBUTING.md. A roadmap for future versions (v0.6+) outlines planned enhancements such as workflow/network promotion and proxy support. The project is currently in a "Beta" status.

Licensing & Compatibility

Licensed under the Apache License 2.0, permitting commercial use and closed-source linking. It is designed for legitimate UI automation, testing, and research, explicitly stating that security or anti-automation challenge completion is not a guaranteed capability. DrissionPage 4.x is supported, but DrissionPage 5.x is explicitly marked as unsupported.

Limitations & Caveats

The project is in "Beta" status, with real browser behavior dependent on the local Chrome/Chromium installation and target websites. DrissionPage 5.x is not supported. The tool does not guarantee success for security or anti-automation challenges.