mcp-rquest

mcp-rquest is a Model Context Protocol (MCP) server that adds advanced HTTP request capabilities to Claude and other LLMs. Built on top of the rquest library, it enables realistic browser-like interactions with websites by emulating TLS fingerprints, JA3/JA4, and HTTP/2 behavior, while supporting a full suite of HTTP methods and flexible authentication. The server also includes utilities for converting HTML and PDF documents into Markdown, making it easier for language models to process fetched content. Core tools expose standard HTTP methods (GET, POST, PUT, DELETE, PATCH, HEAD, OPTIONS, TRACE) and allow you to customize requests with headers, cookies, redirects, query parameters, and various payload formats.

Key capabilities include:

• Comprehensive HTTP methods for interacting with web resources
• Browser fingerprinting to mimic real user clients and bypass bot defenses
• Content handling that counts tokens for large responses and converts HTML/PDF to Markdown via get_stored_response_with_markdown
• PDF support through the Marker-based pipeline for Markdown-ready outputs
• Secure storage of retrieved responses in system temp directories
• Multiple authentication modes (Basic, Bearer, and custom methods)
• Flexible request customization including headers, cookies, redirects, and data formats (form data, JSON, multipart/form-data)

To use, install the package for your environment (see installation steps) and run the server using your preferred runner (uvx, Python module, or pipx as shown in the configuration examples). Once running, you can invoke the available HTTP tools (http_get, http_post, http_put, http_delete, http_patch, http_head, http_options, http_trace) and retrieve or convert responses using the response tools (get_stored_response, get_stored_response_with_markdown, get_model_state, restart_model_loading). The PDF-to-Markdown support is integrated through the same response workflow, enabling seamless processing of documents for LLM consumption.

Prerequisites:

• Python 3.8+ (Recommended)
• pip for Python package installation
• Optional: uv (for fast, user-friendly environment) or uvx if you prefer uv-based workflows

Installation steps (pip):

1. Create and activate a virtual environment (optional but recommended):

python -m venv venv
# Unix/macOS
source venv/bin/activate
# Windows
venv\Scripts\activate

2. Install mcp-rquest from PyPI:

pip install mcp-rquest

3. Run the server as a Python module:

python -m mcp_rquest

Alternative installation methods:

• Using uv (recommended if you want uvx to run the server directly):
  • Install uv and uvx per their documentation, then run the server via uvx with the module name as the entry point (e.g., uvx mcp-rquest) depending on your setup.
• Using pipx:

pipx install mcp-rquest

Then start via:

pipx run mcp-rquest

Configuration usage examples for Claude.app integration are provided in the README and can be adapted to your environment.

Tips and known considerations:

• The server supports automatic HTML-to-Markdown and PDF-to-Markdown conversions; ensure you have the Marker library available or pre-downloaded during installation if you rely on PDF rendering performance.
• When using large responses, monitor token counts and pagination to avoid excessive memory usage; the system stores responses in temporary directories for later retrieval with get_stored_response.
• Ensure you configure appropriate authentication methods (Basic, Bearer, or custom) to protect endpoints when exposing the MCP server in production.
• If you encounter fingerprint or TLS-related issues, verify that the underlying dependencies (rquest and its TLS stack) are up to date and that Python SSL configurations on your host are correct.
• For local development, enable verbose logging to trace HTTP requests, response handling, and Markdown conversion steps to diagnose issues with content transformation.
• The server exposes a complete suite of HTTP methods; when testing, start with GET/HEAD to verify connectivity before advancing to state-changing methods like POST/PUT/PATCH.