duckduckgo

Web Forager is an MCP server that exposes web search and fetch capabilities powered by DuckDuckGo, packaged as a reusable MCP toolset. It provides three primary tools for MCP clients: duckduckgo_search to perform web searches and return results (JSON by default, with an option for plain text), duckduckgo_news_search to fetch recent news articles with source attribution, and web_fetch to retrieve and convert a URL into markdown or JSON (with a direct HTTP fetch attempt first and a Jina Reader fallback if needed). These tools enable integrated knowledge retrieval and content processing within MCP-enabled agents and workflows. To use the server, run it via an MCP client using the configured stdio transport; then call the exposed tools from your client code to perform searches, retrieve news, or fetch and format web pages for downstream processing by your AI models.

Prerequisites:

• Python 3.10 or higher
• uv (recommended) or pip

Install from PyPI (recommended):

# Using uv (preferred)
uv pip install web-forager

# Or using pip
pip install web-forager

Install with UVX (for Claude Desktop):

# Install UVX if you haven't already
pip install uvx

# Install the Web Forager package
uvx install web-forager

Install from source (for development):

# Clone the repository
git clone https://github.com/CyranoB/web-forager.git
cd web-forager

# Install with uv (recommended)
uv pip install -e .

# Or with pip
pip install -e .

Docker deployment (optional):

# Build the image (uses version from latest git tag)
docker build --build-arg VERSION=$(git describe --tags --abbrev=0 | sed 's/^v//') -t web-forager .

# Run the server (MCP clients use STDIO)
docker run -i web-forager

Notes:

• The MCP server runs in STDIO mode and is designed to be used with MCP clients that support the standard tool interface.
• You can customize Python versions via the MCP config (e.g., --python constraints in uvx) to ensure compatibility with your environment.
• The web_fetch tool attempts a direct HTTP fetch first; if that fails or is insufficient, it will fall back to Jina Reader for content extraction.
• When using Docker, ensure your MCP client can connect to the container's STDIO or adapt the entrypoint as needed for your environment.
• If you need to adjust output formats, you can pass output_format parameters to the tools (e.g., json, text, markdown) depending on the MCP client capabilities.
• The server exposes additional internal tools (duckduckgo_search, duckduckgo_news_search, web_fetch) for integration into larger agent workflows.