gemini-search-mcp 0.1.0

MCP server that exposes Gemini web search and document question answering tools.

Gemini Search MCP

Gemini Search MCP packages a Model Context Protocol server that exposes two tools:

• web_search – Uses Gemini with Google Search grounding to answer general questions.
• document_question_answering – Converts local documents to captioned markdown and asks Gemini to answer questions about their contents.

Installation

Python (pip)

pip install gemini-search-mcp

Node.js (npm)

npm install -g gemini-search-mcp

Usage

Set your Google API key (must have Gemini access):

export GOOGLE_API_KEY="your-key"

Run the MCP server (defaults to stdio transport):

gemini-search-mcp run
# or simply
# gemini-search-mcp

Configure Codex automatically (writes to ~/.codex/config.toml by default):

gemini-search-mcp configure --api-key "YOUR_KEY"

Configure Copilot CLI (writes to ~/.copilot/config.json):

gemini-search-mcp configure --cli-type copilot --api-key "YOUR_KEY"

Configure both Codex and Copilot CLI at once:

gemini-search-mcp configure --cli-type both --api-key "YOUR_KEY"

For npm/npx installation with custom command:

gemini-search-mcp configure --command npx --command-args -y gemini-search-mcp --api-key "YOUR_KEY"

Clear cached conversion artifacts:

gemini-search-mcp clear-cache
# 선택 옵션: --cache-dir /custom/path --remove-root

Development

Install in editable mode with testing dependencies:

pip install -e .

Ensure LibreOffice is installed and on PATH if you plan to convert non-PDF documents.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Publishing

For maintainers: See PUBLISHING.md for instructions on how to publish new versions to PyPI and npm.

Changelog

See CHANGELOG.md for a list of changes in each version.

License

MIT – all rights reserved.