MCP Aoai Web Browsing

@kimtth

リポジトリ

基本情報

カテゴリ

ブラウザ自動化

ライセンス

MIT

トランスポート

stdio

作者

kimtth

MCP Aoai Web Browsing について

A minimal Model Context Protocol 🖥️ server/client🧑‍💻with OpenAI and 🌐 web browser control via Playwright.

#mcp #web-automation #playwright #azure-openai #model-context-protocol #mcp-server #mcp-client

MCP Aoai Web Browsing の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "mcp-aoai-web-browsing": {
      "command": "uv",
      "args": [
        "sync"
      ]
    }
  }
}

MCP Aoai Web Browsing のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Mcp Aoai Web Browsing?

Mcp Aoai Web Browsing is a minimal MCP server and client application that combines Playwright-based web browsing with Azure OpenAI or OpenAI. It converts MCP tool responses into the OpenAI function calling format using a custom MCP-LLM Bridge, enabling AI assistants to navigate and interact with web pages programmatically.

How to use Mcp Aoai Web Browsing?

Configure a .env file with your Azure OpenAI endpoint, API key, deployment model, and API version. Install uv, run uv sync to install dependencies, then execute python chatgui.py to launch the chat interface. For external clients (Claude Desktop, VS Code, Claude Code), add a mcp.json configuration pointing to the server script.

Key features of Mcp Aoai Web Browsing

Built with FastMCP and Playwright for web automation
Converts MCP tools to OpenAI function calling format
Supports both Azure OpenAI and standard OpenAI providers
Works in-process or via external stdio MCP server connections
Reusable by different clients (Claude Desktop, VS Code, custom scripts)
Exposes direct tool metadata and execution for custom LLM loops

Use cases of Mcp Aoai Web Browsing

AI agents that need to navigate websites and extract content
Automated web form filling and login workflows
Browser-based testing orchestrated by an LLM

よくある質問

What dependencies are required?

The project uses `uv` for Python library management. During development in December 2024, `pip` and `poetry` were not yet fully supported by the MCP CLI. Playwright is used as the web testing framework.

How do I configure for Azure OpenAI vs standard OpenAI?

For Azure OpenAI, set `AZURE_OPEN_AI_ENDPOINT`, `AZURE_OPEN_AI_API_KEY`, `AZURE_OPEN_AI_DEPLOYMENT_MODEL`, and `AZURE_OPEN_AI_API_VERSION` in `.env`. For standard OpenAI, use the `get_openai_llm_config()` function and set `OPENAI_API_KEY` and `OPENAI_MODEL` environment variables.

What transport layer does the server use?

The server communicates over stdio (a transport layer) using JSON-RPC as the application protocol. This is commonly described as "JSON-RPC over stdio."

How do I connect external clients?

External clients like Claude Desktop, Claude Code, and VS Code connect via `mcp.json` configuration. The configuration specifies the `uv` command to run the FastMCP server script and provides environment variables for Azure OpenAI credentials.

What tools does the server expose?

The server exposes tools such as `playwright_navigate`, which accepts a URL, optional timeout, and wait condition. Tools are defined using FastMCP's `@self.mcp.tool()` decorator, and their docstrings provide descriptions used for function calling.