# Gemini API Docs

A local stdio MCP server that keeps a copy of the Gemini API docs in SQLite so your agent can search them offline.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: kgprs
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/kgprs/gemini-api-docs-mcp
- Updated: 2026-10-29
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/mcp/gemini-api-docs

## What it does

The server answers from a local index instead of the web. On startup it fetches the page list from `https://ai.google.dev/gemini-api/docs/llms.txt`, downloads every page, strips the text and writes it into SQLite with an FTS5 index. Lookups hit that database.

Three tools ship with it. `search_documentation` takes up to three queries per call, `get_capability_page` returns the section list or the body of one section, and `get_current_model` serves the page about the current Gemini models. Ingestion starts right away, so the first launch is slow and the server answers nothing until it finishes.

The database lives at `~/.mcp/gemini-api-docs/database.db` unless `GEMINI_DOCS_DB_PATH` points elsewhere. Going through `uvx` skips installation entirely. With the `pip` route you start `gemini-docs-mcp` yourself afterwards, otherwise the server never comes up.

## Requirements
- Needs: Docker

## Install

### Claude Code (native platform)

**Install**
- The MCP registry has no launch command for it: see the repository README

## When to use it
- look up the embeddings example in the Gemini API docs
- check which Gemini model is current without leaving the editor
- list the doc sections and pull the text of one page
- keep a local docs index that refreshes on every startup
- have the agent cite the official page instead of guessing

## Example prompts
- Search the Gemini API documentation for how request batching works
- Show me the current Gemini model page from the docs server
- List the capability pages and open the one on embeddings

## Similar
- [gemini-api-dev](https://affpapa.org/en/nevibecode/skills/gemini-api-dev.md) — skill · 27800 installs
- [gemini-agents-api](https://affpapa.org/en/nevibecode/skills/gemini-agents-api.md) — skill · 9466 installs
- [gemini-api](https://affpapa.org/en/nevibecode/skills/gemini-api.md) — skill · 17116 installs
- [gemini-live-api-dev](https://affpapa.org/en/nevibecode/skills/gemini-live-api-dev.md) — skill · 11300 installs
- [StitchAPI Docs](https://affpapa.org/en/nevibecode/mcp/rejifald-stitchapi-docs.md) — MCP server · 722 installs
- [coralogix-docs](https://affpapa.org/en/nevibecode/skills/coralogix-docs.md) — skill · 5000 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
