Gemini Docs MCP Server
# Gemini Docs MCP Server
An MCP (Model Context Protocol) server that provides tools to search and fetch Google's Gemini API documentation.
## Features
- **Search Documentation**: Search through the Gemini API documentation index to find relevant pages
- **Fetch Documentation**: Fetch and parse specific documentation pages with structured content extraction
## Installation
```bash
npm install
npm run build
```
## Usage
### As an MCP Server (stdio)
```bash
node dist/index.js
```
### Configuration for Claude Desktop
Add to your Claude Desktop config (`~/Library/Application Support/Claude/claude_desktop_config.json`):
```json
{
"mcpServers": {
"gemini-docs": {
"command": "node",
"args": ["/path/to/gemini-docs-mcp/dist/index.js"]
}
}
}
```
## Tools
### `search_gemini_docs`
Search the Gemini API documentation for relevant pages.
**Parameters:**
- `query` (string, required): Search query (e.g., "function calling", "embeddings")
- `max_results` (number, optional): Maximum results to return (1-20, default: 10)
- `response_format` (string, optional): Output format - "markdown" or "json" (default: "markdown")
**Example:**
```json
{
"query": "function calling",
"max_results": 5,
"response_format": "json"
}
```
### `fetch_gemini_doc`
Fetch and parse a specific Gemini API documentation page.
**Parameters:**
- `path` (string, required): Documentation path (e.g., "embeddings", "function-calling"). Use empty string for the main overview page.
- `response_format` (string, optional): Output format - "markdown" or "json" (default: "markdown")
**Common paths:**
- `""` - Main overview page
- `quickstart` - Getting started guide
- `function-calling` - Function calling / tool use
- `embeddings` - Text embeddings
- `structured-output` - JSON structured output
- `text-generation` - Text generation basics
- `image-understanding` - Vision / image analysis
- `live` - Live API (real-time streaming)
- `api-key` - API key setup
- `models` - Available models
**Example:**
```json
{
"path": "function-calling",
"response_format": "markdown"
}
```
## Development
```bash
# Install dependencies
npm install
# Development mode with auto-reload
npm run dev
# Build
npm run build
# Run tests
npm run test
```
TDQS
Scored across 2 tools
The two tools have clearly distinct purposes with no overlap: fetch_gemini_doc retrieves and parses specific documentation pages, while search_gemini_docs searches across the documentation index to find relevant pages. The descriptions explicitly guide users to use search first when unsure of the path, then fetch for detailed content, creating a logical workflow without confusion.
Both tools follow a perfect verb_noun pattern with consistent snake_case naming: fetch_gemini_doc and search_gemini_docs. The naming clearly indicates the action (fetch/search) and the target (gemini_doc/gemini_docs), making them predictable and easy to understand as a set.
With only 2 tools, this server feels thin for a documentation exploration domain. While the tools cover the core workflow of search-then-fetch, additional operations like listing available documentation categories or getting metadata about the documentation set would enhance completeness. The count is borderline minimal but functional.
For a documentation server, the surface covers the essential workflow: searching for relevant pages and fetching detailed content. The main gap is the lack of a tool to browse or list available documentation sections without searching, which could help agents discover content. However, the search tool with its comprehensive examples mitigates this gap reasonably well.