python -client

This MCP server implements a Python-based MCP server/client setup that enables large language models to access structured documents and web content through a unified MCP interface. The server exposes a tool-enabled flow that can fetch and aggregate documentation from various libraries (e.g., LangChain, llama-index, AutoGen, and MCP docs) by querying sites and extracting text for use in downstream reasoning. Clients can connect via the Stdio transport (local usage) or via SSE for remote deployments, and can be controlled through the MCP runtime using common commands such as starting the server with uv run main.py. Tools are exposed via a get_docs function that takes a query and a library, validates the library, performs a site-specific search, and returns extracted text from relevant pages. This enables the model to reference up-to-date documentation and API surfaces when answering user questions.

Prerequisites:

• Python 3.8+ installed on your system
• uv (UvS or UV CLI) installed or available in PATH
• Basic familiarity with Python packaging and virtual environments

1. Install UV (if not already installed):

• macOS/Linux: curl -LsSf https://astral.sh/uv/install.sh | sh
• Windows: powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

2. Create and activate a project directory for the MCP server client:

• uv init gobinfan-python-mcp-server-client
• cd gobinfan-python-mcp-server-client
• uv venv
• source .venv/bin/activate # Windows: .venv\Scripts\activate

3. Install MCP CLI and dependencies:

• uv add "mcp[cli]" httpx

4. Prepare your server script (main.py) in the project root. This repository provides the MCP server/client code as described in the README. Ensure main.py contains the MCP server setup (stdio transport) and exposes the get_docs tool if using the provided example.

5. Run the server:

• uv run main.py

Note: If you prefer using a boilerplate without UV, you can adapt the server to a standard Python environment by installing required dependencies from the project (for example, pip install httpx beautifulsoup4). The MCP server will listen for commands via the configured transport (stdio by default in the examples).

Tips and common considerations:

• For local development, using the stdio transport is simplest. For cloud or remote deployments, consider SSE transport and a compatible hosting setup.
• You may need an API key for SERPER (Google search) if you enable the web-doc tooling (SERPER_API_KEY in environment). Ensure to set SERPER_API_KEY in your environment when running the server if you enable doc searching.
• The get_docs tool relies on network access to fetch and parse documentation pages; occasional timeouts may occur due to network or site protections. The sample code handles timeouts gracefully.
• If you modify docs URLs or libraries, update the docs_urls mapping accordingly and ensure the library key exists in the validation step.
• When configuring MCP clients (e.g., Cline or Cursor), provide the correct directory path to your project and ensure the Python script (main.py) can be launched with the specified command sequence.