@houtini/gemini-mcp 1.4.2 → 2.2.0

package/README.md

@@ -1,784 +1,314 @@
-# Gemini MCP Server
-
-[![npm version](https://img.shields.io/npm/v/@houtini/gemini-mcp.svg?style=flat-square)](https://www.npmjs.com/package/@houtini/gemini-mcp)
-[![MCP Registry](https://img.shields.io/badge/MCP-Registry-blue?style=flat-square)](https://registry.modelcontextprotocol.io)
-[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg?style=flat-square)](https://opensource.org/licenses/Apache-2.0)
-[![TypeScript](https://img.shields.io/badge/TypeScript-5.3-blue?style=flat-square&logo=typescript)](https://www.typescriptlang.org/)
-[![MCP](https://img.shields.io/badge/MCP-Compatible-green?style=flat-square)](https://modelcontextprotocol.io)
-
-A production-ready Model Context Protocol server for Google's Gemini AI models. I've built this with TypeScript and the latest MCP SDK (1.25.3), focusing on real-world reliability rather than feature bloat.
-
-## What This Does
-
-This server connects Claude Desktop (or any MCP client) to Google's Gemini models. The integration is straightforward: chat with Gemini, get model information, and run deep research tasks with Google Search grounding built in.
-
-What I think matters here: the server discovers available models automatically from Google's API, which means you're always working with the latest releases without updating configuration files. No hardcoded model lists that go stale.
-
-## Quick Start
-
-The simplest way to use this is with `npx` - no installation required:
-
-```bash
-# Get your API key from Google AI Studio first
-# https://makersuite.google.com/app/apikey
-
-# Test it works (optional)
-npx @houtini/gemini-mcp
-
-# Add to Claude Desktop (configuration below)
-```
-
-## Installation Options
-
-### Recommended: npx (No Installation)
-
-```bash
-npx @houtini/gemini-mcp
-```
-
-This approach pulls the latest version automatically. I prefer this because you don't clutter your system with global packages, and updates happen transparently.
-
-### Alternative: Global Installation
-
-```bash
-npm install -g @houtini/gemini-mcp
-gemini-mcp
-```
-
-### Alternative: Local Project
-
-```bash
-npm install @houtini/gemini-mcp
-npx @houtini/gemini-mcp
-```
-
-### From Source (Developers)
-
-```bash
-git clone https://github.com/houtini-ai/gemini-mcp.git
-cd gemini-mcp
-npm install
-npm run build
-npm start
-```
-
-## Configuration
-
-### Step 1: Get Your API Key
-
-Visit [Google AI Studio](https://makersuite.google.com/app/apikey) to create a free API key. This takes about 30 seconds.
-
-### Step 2: Configure Claude Desktop
-
-Add this to your Claude Desktop config file:
-
-**Windows**: `%APPDATA%\Claude\claude_desktop_config.json`  
-**macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
-
-#### Using npx (Recommended)
-
-```json
-{
-  "mcpServers": {
-    "gemini": {
-      "command": "npx",
-      "args": ["@houtini/gemini-mcp"],
-      "env": {
-        "GEMINI_API_KEY": "your-api-key-here"
-      }
-    }
-  }
-}
-```
-
-#### Using Global Installation
-
-```json
-{
-  "mcpServers": {
-    "gemini": {
-      "command": "gemini-mcp",
-      "env": {
-        "GEMINI_API_KEY": "your-api-key-here"
-      }
-    }
-  }
-}
-```
-
-Requires `npm install -g @houtini/gemini-mcp` first.
-
-#### Using Local Build
-
-```json
-{
-  "mcpServers": {
-    "gemini": {
-      "command": "node",
-      "args": ["./node_modules/@houtini/gemini-mcp/dist/index.js"],
-      "env": {
-        "GEMINI_API_KEY": "your-api-key-here"
-      }
-    }
-  }
-}
-```
-
-Only works if installed locally in the current directory.
-
-### Step 3: Restart Claude Desktop
-
-After updating the config, restart Claude Desktop. The server loads on startup.
-
-### Optional: Additional Configuration
-
-```json
-{
-  "mcpServers": {
-    "gemini": {
-      "command": "npx",
-      "args": ["@houtini/gemini-mcp"],
-      "env": {
-        "GEMINI_API_KEY": "your-api-key-here",
-        "LOG_LEVEL": "info",
-        "GEMINI_ALLOW_EXPERIMENTAL": "false"
-      }
-    }
-  }
-}
-```
-
-**Environment Variables:**
-
-| Variable | Default | What It Does |
-|----------|---------|--------------|
-| `GEMINI_API_KEY` | *required* | Your Google AI Studio API key |
-| `LOG_LEVEL` | `info` | Logging detail: `debug`, `info`, `warn`, `error` |
-| `GEMINI_ALLOW_EXPERIMENTAL` | `false` | Include experimental models (set `true` to enable) |
-
-## Dynamic Model Discovery
-
-The server automatically discovers available Gemini models from Google's API on first use. This happens transparently - you don't need to configure anything.
-
-### How It Works
-
-1. Server starts instantly with reliable fallback models
-2. First request triggers model discovery from Google's API (adds 1-2 seconds once)
-3. Subsequent requests use the discovered models (no delay)
-4. If discovery fails, fallback models work immediately
-
-What I've found: this approach keeps you current with Google's releases whilst maintaining instant startup. The server filters to stable production models by default, which avoids experimental model rate limits.
-
-### What Gets Discovered
-
-- All available Gemini models (stable and experimental)
-- Accurate context window sizes directly from Google
-- Model capabilities and recommended use cases
-- Latest releases as soon as Google makes them available
-
-The default model selection prioritises: stable models over experimental, newest version available, Flash variants for speed, and capability matching for your request type.
-
-### Performance Impact
-
-- Startup: 0ms (instant)
-- First request: +1-2 seconds (one-time discovery)
-- Subsequent requests: 0ms overhead
-- Discovery failure: 0ms (uses fallback immediately)
-
-Check your logs after first request to see what was discovered:
-```
-Models discovered from API (count: 38, defaultModel: gemini-2.5-flash)
-```
-
-## Experimental Models
-
-By default, the server uses stable production models. This ensures reliable performance and avoids Google's stricter rate limits on experimental releases.
-
-### Stable vs Experimental
-
-**Stable Models** (default behaviour):
-- Production-ready
-- Better rate limits
-- Consistent performance
-- Examples: `gemini-2.5-flash`, `gemini-2.5-pro`, `gemini-2.0-flash`
-
-**Experimental Models** (opt-in):
-- Latest features before stable release
-- Stricter rate limits
-- Potentially unexpected behaviour
-- Can be deprecated quickly
-- Examples: `gemini-exp-1206`, `gemini-2.0-flash-thinking-exp`
-
-### Enabling Experimental Models
-
-Set `GEMINI_ALLOW_EXPERIMENTAL=true` in your configuration:
-
-```json
-{
-  "mcpServers": {
-    "gemini": {
-      "command": "npx",
-      "args": ["@houtini/gemini-mcp"],
-      "env": {
-        "GEMINI_API_KEY": "your-api-key-here",
-        "GEMINI_ALLOW_EXPERIMENTAL": "true"
-      }
-    }
-  }
-}
-```
-
-This includes experimental models in discovery and makes them eligible as defaults. You can still explicitly request any model regardless of this setting - the flag only affects which models are used automatically.
-
-### When to Enable
-
-Keep experimental disabled if you need reliable, consistent performance or you're building production applications.
-
-Enable experimental if you're testing cutting-edge features, doing research, or you understand the rate limit trade-offs.
-
-## Usage Examples
-
-### Basic Chat
-
-```
-Can you help me understand quantum computing using Gemini?
-```
-
-Claude automatically uses the `gemini_chat` tool.
-
-### Creative Writing
-
-```
-Use Gemini to write a short story about artificial intelligence discovering creativity.
-```
-
-### Technical Analysis
-
-```
-Use Gemini Pro to explain the differences between various machine learning algorithms.
-```
-
-### Model Selection
-
-```
-Use Gemini 1.5 Pro to analyse this code and suggest improvements.
-```
-
-### Getting Model Information
-
-```
-Show me all available Gemini models and their capabilities.
-```
-
----
-
-## Complete Prompting Guide
-
-Check the **[Comprehensive Prompting Guide](PROMPTING_GUIDE.md)** for:
-
-- Advanced prompting techniques
-- Model selection strategies
-- Parameter tuning (temperature, tokens, system prompts)
-- Using Google Search grounding
-- Creative workflows and use cases
-- Best practices
-- Troubleshooting
-
-**[Read the Prompting Guide](PROMPTING_GUIDE.md)**
-
----
-
-## Google Search Grounding
-
-Google Search grounding is built in and enabled by default. This gives Gemini models access to current web information, which significantly improves accuracy for questions requiring up-to-date data.
-
-### What It Does
-
-When you ask a question that benefits from current information:
-1. Analyses your query to determine if web search helps
-2. Generates relevant search queries automatically
-3. Performs Google searches using targeted queries
-4. Processes results and synthesises information
-5. Provides enhanced response with inline citations
-6. Shows search metadata including queries used
-
-### Best Use Cases
-
-**Current Events & News**
-```
-What are the latest developments in AI announced this month?
-Recent breakthroughs in quantum computing research?
-```
-
-**Real-time Data**
-```
-Current stock prices for major tech companies
-Today's weather forecast for London
-```
-
-**Recent Developments**
-```
-New software releases this week
-Latest scientific discoveries in medicine
-```
-
-**Fact Checking**
-```
-Verify recent statements about climate change
-Check the latest statistics on global internet usage
-```
-
-### Controlling Grounding
-
-Grounding is enabled by default. Disable it for purely creative or hypothetical responses:
-
-```
-Use Gemini without web search to write a fictional story about dragons in space.
-```
-
-For API calls, use the `grounding` parameter:
-
-```json
-{
-  "message": "Write a creative story about time travel",
-  "grounding": false
-}
-```
-
-### Understanding Grounded Responses
-
-Grounded responses include source citations and search transparency:
-
-```
-Sources: (https://example.com/article1) (https://example.com/article2)
-Search queries used: latest AI developments 2025, OpenAI GPT-5 release
-```
-
-What I've found: grounding dramatically reduces hallucinations for factual queries whilst maintaining creative flexibility when you need it.
-
-## Deep Research
-
-The server includes deep research capability that performs iterative multi-step research on complex topics. This synthesises comprehensive reports with proper citations.
-
-### How It Works
-
-Deep research conducts multiple research iterations:
-
-1. Initial broad exploration
-2. Gap analysis identifying what's missing
-3. Targeted research into specific areas
-4. Synthesis into comprehensive report
-5. Iteration until thorough coverage
-
-### Using Deep Research
-
-```
-Use Gemini deep research to investigate the impact of quantum computing on cybersecurity.
-```
-
-With parameters:
-```
-Use Gemini deep research with 7 iterations to create a comprehensive report on renewable energy trends, focusing on solar and wind power adoption rates.
-```
-
-### Research Parameters
-
-| Parameter | Type | Default | What It Does |
-|-----------|------|---------|--------------|
-| `research_question` | string | *required* | The topic to investigate |
-| `max_iterations` | integer | 5 | Research cycles (3-10) |
-| `focus_areas` | array | - | Specific aspects to emphasise |
-| `model` | string | *latest stable* | Which model to use |
-
-### Best For
-
-- Academic research and literature reviews
-- Market analysis and competitive intelligence
-- Technology trend analysis
-- Policy research and impact assessments
-- Multi-faceted business problems
-
-### Configuring Iterations by Environment
-
-Different AI environments have different timeout tolerances:
-
-**Claude Desktop (3-5 iterations recommended)**
-- Timeout: ~4 minutes
-- Safe maximum: 5 iterations
-- Use 3-4 for most tasks
-
-**Agent SDK / IDEs (7-10 iterations recommended)**
-- Timeout: 10+ minutes
-- Maximum: 10 iterations
-- Use 7-10 for comprehensive research
-
-**AI Platforms like Cline, Roo-Cline (7-10 iterations)**
-- Similar to Agent SDK
-- Can handle longer processes
-
-### Handling Timeouts
-
-If you hit timeout or thread limits:
-
-1. Reduce iterations (start with 3)
-2. Narrow focus using `focus_areas` parameter
-3. Split complex topics into smaller research tasks
-4. Check which environment you're using
-
-Example with focused research:
-```
-Use Gemini deep research with 3 iterations focusing on cost analysis and market adoption to examine solar panel technology trends.
-```
-
-Deep research takes several minutes. It's designed for comprehensive analysis rather than quick answers.
-
-## API Reference
-
-### gemini_chat
-
-Chat with Gemini models.
-
-**Parameters:**
-
-| Parameter | Type | Required | Default | What It Does |
-|-----------|------|----------|---------|--------------|
-| `message` | string | Yes | - | The message to send |
-| `model` | string | No | *Latest stable* | Which model to use |
-| `temperature` | number | No | 0.7 | Randomness (0.0-1.0) |
-| `max_tokens` | integer | No | 8192 | Maximum response length (1-32768) |
-| `system_prompt` | string | No | - | System instruction |
-| `grounding` | boolean | No | true | Enable Google Search |
-
-**Example:**
-```json
-{
-  "message": "What are the latest developments in quantum computing?",
-  "model": "gemini-1.5-pro",
-  "temperature": 0.5,
-  "max_tokens": 1000,
-  "system_prompt": "You are a technology expert. Provide current information with sources.",
-  "grounding": true
-}
-```
-
-### gemini_list_models
-
-Retrieve information about discovered Gemini models.
-
-**Parameters:** None required
-
-**Example:**
-```json
-{}
-```
-
-**Response includes:**
-- Model names and display names
-- Descriptions of strengths
-- Context window sizes from Google
-- Recommended use cases
-
-### gemini_deep_research
-
-Conduct iterative multi-step research.
-
-**Parameters:**
-
-| Parameter | Type | Required | Default | What It Does |
-|-----------|------|----------|---------|--------------|
-| `research_question` | string | Yes | - | Topic to research |
-| `max_iterations` | integer | No | 5 | Research cycles (3-10) |
-| `focus_areas` | array | No | - | Specific areas to emphasise |
-| `model` | string | No | *Latest stable* | Model to use |
-
-**Example:**
-```json
-{
-  "research_question": "Impact of AI on healthcare diagnostics",
-  "max_iterations": 7,
-  "focus_areas": ["accuracy improvements", "cost implications", "regulatory challenges"]
-}
-```
-
-### Available Models
-
-Models are dynamically discovered from Google's API. Typical available models:
-
-| Model | Best For | Description |
-|-------|----------|-------------|
-| **gemini-2.5-flash** | General use | Latest Flash - fast, versatile |
-| **gemini-2.5-pro** | Complex reasoning | Latest Pro - advanced capabilities |
-| **gemini-2.0-flash** | Speed-optimised | Gemini 2.0 Flash - efficient |
-| **gemini-1.5-flash** | Quick responses | Gemini 1.5 Flash - fast |
-| **gemini-1.5-pro** | Large context | 2M token context window |
-
-Use `gemini_list_models` to see exact available models with current context limits.
-
-## Development
-
-### Building from Source
-
-```bash
-git clone https://github.com/houtini-ai/gemini-mcp.git
-cd gemini-mcp
-npm install
-npm run build
-npm run dev
-```
-
-### Scripts
-
-| Command | What It Does |
-|---------|--------------|
-| `npm run build` | Compile TypeScript |
-| `npm run dev` | Development mode with live reload |
-| `npm start` | Run compiled server |
-| `npm test` | Run tests |
-| `npm run lint` | Check code style |
-| `npm run lint:fix` | Fix linting issues |
-
-### Project Structure
-
-```
-src/
-├── config/           # Configuration management
-├── services/         # Business logic
-│   └── gemini/       # Gemini API integration
-├── tools/            # MCP tool implementations
-├── utils/            # Logger and error handling
-├── cli.ts            # CLI entry
-└── index.ts          # Main server
-```
-
-### Architecture
-
-The server follows clean, layered architecture:
-
-1. CLI Layer - Command-line interface
-2. Server Layer - MCP protocol handling
-3. Tools Layer - MCP tool implementations
-4. Service Layer - Business logic and API integration
-5. Utility Layer - Logging and error handling
-
-## Troubleshooting
-
-### "GEMINI_API_KEY environment variable not set"
-
-Check your Claude Desktop configuration includes the API key in the `env` section.
-
-### Server Not Appearing in Claude Desktop
-
-1. Restart Claude Desktop after configuration changes
-2. Verify config file path:
-   - Windows: `%APPDATA%\Claude\claude_desktop_config.json`
-   - macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
-3. Validate JSON syntax
-4. Test your API key at [Google AI Studio](https://makersuite.google.com/app/apikey)
-
-### "Module not found" with npx
-
-```bash
-# Clear npx cache
-npx --yes @houtini/gemini-mcp
-
-# Or install globally
-npm install -g @houtini/gemini-mcp
-```
-
-### Node.js Version Issues
-
-```bash
-# Check version
-node --version
-
-# Should be v18.0.0 or higher
-# Update from https://nodejs.org
-```
-
-### Debug Mode
-
-Enable detailed logging:
-
-```json
-{
-  "mcpServers": {
-    "gemini": {
-      "command": "npx",
-      "args": ["@houtini/gemini-mcp"],
-      "env": {
-        "GEMINI_API_KEY": "your-api-key-here",
-        "LOG_LEVEL": "debug"
-      }
-    }
-  }
-}
-```
-
-### Log Files
-
-Logs are written to:
-- Console output (Claude Desktop developer tools)
-- `logs/combined.log` - All levels
-- `logs/error.log` - Errors only
-
-### Testing Your Setup
-
-Test with these queries:
-1. "Can you list the available Gemini models?"
-2. "Use Gemini to explain photosynthesis."
-3. "Use Gemini 1.5 Pro with temperature 0.9 to write a creative poem about coding."
-
-### Performance Tuning
-
-For better performance:
-
-- Adjust token limits based on your use case
-- Use appropriate models (Flash for speed, Pro for complexity)
-- Monitor logs for rate limiting issues
-- Set temperature values appropriately (0.7 balanced, 0.3 focused, 0.9 creative)
-
-## Contributing
-
-Contributions welcome. Follow these steps:
-
-1. Fork the repository
-2. Create a feature branch: `git checkout -b feature/amazing-feature`
-3. Make your changes and add tests
-4. Run tests: `npm test`
-5. Lint: `npm run lint:fix`
-6. Build: `npm run build`
-7. Commit: `git commit -m 'Add amazing feature'`
-8. Push: `git push origin feature/amazing-feature`
-9. Open a Pull Request
-
-### Development Guidelines
-
-- Follow TypeScript best practices
-- Add tests for new functionality
-- Update documentation
-- Use conventional commit messages
-- Maintain backwards compatibility
-
-## Technical Details
-
-### Migration to MCP SDK 1.25.3
-
-This server has been migrated to the latest MCP SDK (1.25.3) with ES modules support. Key technical changes:
-
-**SDK Updates:**
-- Migrated from `Server` class to `McpServer` API
-- Tool registration uses `registerTool` with Zod validation
-- ES modules throughout (`"type": "module"`)
-- TypeScript configured for `nodenext` module resolution
-
-**Compatibility:**
-- Node.js 18+ (changed from 24+ for broader compatibility)
-- All imports use `.js` extensions for ES module compliance
-- Zod schemas for runtime type validation
-- Modern MCP protocol implementation
-
-**Build System:**
-- TypeScript compiles to ES2022 modules
-- Clean separation between business logic and MCP interface
-- Preserved all Gemini API client functionality
-
-What this means practically: the server now follows modern Node.js and MCP standards, which should prevent compatibility issues with future Claude Desktop updates whilst maintaining all existing functionality.
-
-## Licence
-
-This project is licensed under the Apache 2.0 Licence - see the [LICENSE](LICENSE) file for details.
-
-## Disclaimer
-
-**Use at Your Own Risk**: This software is provided "as is" without warranty. The authors accept no responsibility for damages, data loss, or other issues arising from use.
-
-**Content Safety**: This server interfaces with Google's Gemini AI models. Whilst content safety settings are implemented, AI-generated content quality cannot be guaranteed. Users are responsible for reviewing AI output before use and ensuring compliance with applicable laws.
-
-**API Key Security**: Your Google Gemini API key is sensitive. Keep it confidential, don't commit it to version control, rotate if exposed, and manage API usage costs.
-
-**Data Privacy**: This server processes data through the Model Context Protocol. Avoid sending sensitive or confidential information. Review Google's privacy policy and implement appropriate data handling.
-
-**Production Use**: Users deploying in production should conduct security audits, implement monitoring, have incident response procedures, and regularly update dependencies.
-
-**Third-Party Services**: This software relies on external services (Google Gemini API, npm packages). Service availability, pricing, and functionality may change.
-
-**No Professional Advice**: AI-generated content should not be considered professional advice (legal, medical, financial) without verification by qualified professionals.
-
-By using this software, you acknowledge these terms and agree to use at your own risk.
-
-## Support
-
-- **GitHub Issues**: [Report bugs or request features](https://github.com/houtini-ai/gemini-mcp/issues)
-- **GitHub Discussions**: [Ask questions or share ideas](https://github.com/houtini-ai/gemini-mcp/discussions)
-
-## Changelog
-
-### v1.3.2 - Node.js 18+ Compatibility & Modern SDK
-
-**Breaking Changes:** None (all tool interfaces preserved)
-
-**Technical Updates:**
-- Updated to MCP SDK 1.25.3 (from 1.19.1)
-- Migrated to ES modules (`"type": "module"`)
-- Changed Node.js requirement to >=18.0.0 (from >=24.0.0) for broader compatibility
-- Migrated from `Server` to `McpServer` API
-- Implemented Zod schema validation for all tools
-- Updated TypeScript config to `nodenext` module resolution
-
-**Fixes:**
-- Resolved Node.js v24 ERR_MODULE_NOT_FOUND errors
-- Fixed TypeScript compilation with DOM types for fetch API
-- All imports now use `.js` extensions for ES module compliance
-
-**What This Means:**
-The server now works reliably with Node.js 18, 20, 22, and 24. All existing functionality preserved - this is purely a technical infrastructure update for better compatibility.
-
-### v1.1.0 - Deep Research & Enhanced Discovery
-
-**New Features:**
-- Added deep research capability for iterative analysis
-- Enhanced model discovery with better filtering
-- Improved default model selection logic
-- Better handling of experimental vs stable models
-
-### v1.0.4 - Security & Dependencies
-
-**Updates:**
-- Updated @google/generative-ai to v0.24.1
-- Updated @modelcontextprotocol/sdk to v1.19.1
-- Changed safety settings to BLOCK_MEDIUM_AND_ABOVE
-- Added comprehensive disclaimer
-- Zero vulnerabilities in dependencies
-
-### v1.0.3 - Enhanced Grounding
-
-**Improvements:**
-- Fixed grounding metadata field names
-- Enhanced source citation processing
-- Improved grounding reliability
-- Better error handling for grounding
-
-### v1.0.2 - Google Search Grounding
-
-**New Features:**
-- Added Google Search grounding (enabled by default)
-- Real-time web search integration
-- Source citations in responses
-- Configurable grounding parameter
-
-### v1.0.0 - Initial Release
-
-**Core Features:**
-- Complete TypeScript rewrite
-- Professional modular architecture
-- Comprehensive error handling
-- Full MCP protocol compliance
-- Multiple Gemini model support
-- NPM package distribution
-- Production-ready build system
-
----
-
-**Built for the Model Context Protocol community**
-
-For more about MCP, visit [modelcontextprotocol.io](https://modelcontextprotocol.io)
+# @houtini/gemini-mcp
+
+[![npm version](https://img.shields.io/npm/v/@houtini/gemini-mcp.svg?style=flat-square)](https://www.npmjs.com/package/@houtini/gemini-mcp)
+[![MCP Registry](https://img.shields.io/badge/MCP-Registry-blue?style=flat-square)](https://registry.modelcontextprotocol.io)
+
+**I've been running this MCP server in my Claude Desktop setup for several months, and it's one of the few I leave enabled permanently.** Not because Gemini replaces Claude -- it doesn't -- but because grounded search, deep research, image generation, and video are things Gemini does well. Having them as tools inside Claude beats switching between browser tabs.
+
+Thirteen tools. One `npx` command.
+
+### MCP App previews
+
+Generated images and diagrams render inline in Claude Desktop with zoom controls, file paths, and prompt context:
+
+| Image generation | SVG / diagram generation |
+|:---:|:---:|
+| ![Image preview in MCP App](image-preview-mcp-app.jpg) | ![Diagram preview in MCP App](diagram-preview-mcp-app.jpg) |
+
+---
+
+## Get started in two minutes
+
+**Step 1: Get a Gemini API key**
+
+Go to [Google AI Studio](https://aistudio.google.com/apikey) and create one. The free tier covers most development use -- you'll hit rate limits on deep research if you're hammering it, but for day-to-day work it's fine.
+
+**Step 2: Add to your Claude Desktop config**
+
+Config file locations:
+- Windows: `C:\Users\{username}\AppData\Roaming\Claude\claude_desktop_config.json`
+- macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`
+
+```json
+{
+  "mcpServers": {
+    "gemini": {
+      "command": "npx",
+      "args": ["@houtini/gemini-mcp"],
+      "env": {
+        "GEMINI_API_KEY": "your-api-key-here"
+      }
+    }
+  }
+}
+```
+
+**Step 3: Restart Claude Desktop**
+
+That's it. The tools show up automatically. `npx` pulls the package on first run -- no separate install.
+
+### Local build instead
+
+For development, or if you'd rather not rely on npx:
+
+```bash
+git clone https://github.com/houtini-ai/gemini-mcp
+cd gemini-mcp
+npm install --include=dev
+npm run build
+```
+
+Then point your config at the local build:
+
+```json
+{
+  "mcpServers": {
+    "gemini": {
+      "command": "node",
+      "args": ["C:/path/to/gemini-mcp/dist/index.js"],
+      "env": {
+        "GEMINI_API_KEY": "your-api-key-here"
+      }
+    }
+  }
+}
+```
+
+---
+
+## What it does
+
+### Chat with Google Search grounding
+
+```
+Use gemini:gemini_chat to ask: "What changed in the MCP spec in the last month?"
+```
+
+Grounding is on by default. Gemini searches Google before answering, so you get current information rather than training data cutoff answers. Sources come back as markdown links.
+
+For questions where you want reasoning over live search -- "explain this code" or similar -- set `grounding: false`.
+
+Supports `thinking_level` on Gemini 3 models: `high` for maximum reasoning depth, `low` to keep it fast, `medium`/`minimal` on Gemini 3 Flash only.
+
+### Deep research
+
+```
+Use gemini:gemini_deep_research with:
+  research_question="What are the current approaches to AI agent memory management?"
+  max_iterations=5
+```
+
+Runs multiple grounded search iterations, then synthesises a full report. Takes 2-5 minutes depending on complexity. Worth it for anything where you need comprehensive coverage rather than a quick answer.
+
+Set `max_iterations` to 3-4 in Claude Desktop (4-minute tool timeout). In IDEs (Cursor, Windsurf, VS Code) or agent frameworks with longer timeout tolerance, 7-10 iterations produces noticeably better synthesis. Pass `focus_areas` as an array to steer toward specific angles.
+
+### Image generation with search grounding
+
+```
+Use gemini:generate_image with:
+  prompt="Stock price chart showing Apple (AAPL) closing prices for the last 5 trading days"
+  use_search=true
+  aspectRatio="16:9"
+```
+
+Default model is `gemini-3-pro-image-preview` (Nano Banana Pro). Also supports `gemini-2.5-flash-image` for faster generation.
+
+When `use_search=true`, Gemini searches Google for current data before generating. Financial and news queries work reliably and return 2-5 grounding sources as markdown links. Weather queries are inconsistent (Gemini API limitation, not a code issue).
+
+### Video generation with Veo 3.1
+
+```
+Use gemini:generate_video with:
+  prompt="A close-up shot of a futuristic coffee machine brewing a glowing blue espresso, steam rising dramatically. Cinematic lighting."
+  resolution="1080p"
+  durationSeconds=8
+```
+
+Uses Google's Veo 3.1 model. Generates 4-8 second videos at up to 4K resolution with native synchronised audio. Processing takes 2-5 minutes -- the tool polls automatically until the video is ready.
+
+Options worth knowing about:
+- `aspectRatio` -- `16:9` (landscape, default) or `9:16` (portrait/vertical)
+- `generateAudio` -- on by default, produces dialogue and sound effects matching the prompt
+- `sampleCount` -- generate up to 4 variations in one call
+- `seed` -- for deterministic output across runs
+- `generateThumbnail` -- extracts a frame via ffmpeg (needs ffmpeg in PATH)
+- `generateHTMLPlayer` -- creates a local HTML player alongside the video
+
+### SVG generation
+
+```
+Use gemini:generate_svg with:
+  prompt="Architecture diagram showing a microservices system with API gateway, three services, and a shared database"
+  style="technical"
+  width=1000
+  height=600
+```
+
+Generates clean, production-ready SVG code for diagrams, illustrations, icons, and data visualisations. Styles: `technical` (diagrams), `artistic` (illustrations), `minimal` (simple), `data-viz` (charts).
+
+### Image editing and analysis
+
+**Conversational editing** -- Gemini 3 Pro Image maintains context across editing turns using thought signatures. The server captures these automatically. Pass them back on subsequent edit calls for full continuity:
+
+```
+Use gemini:edit_image with:
+  prompt="Change the colour scheme to blue and green"
+  images=[{data: imageBase64, mimeType: "image/png", thoughtSignature: "fromPreviousCall"}]
+```
+
+Skip thought signatures and each edit starts from scratch.
+
+**Analysis** -- two tools for different purposes:
+- `describe_image` -- Fast general descriptions using Gemini 3 Flash
+- `analyze_image` -- Structured extraction and detailed reasoning using Gemini 3.1 Pro
+
+**Load local files:**
+```
+Use gemini:load_image_from_path with filePath="C:/screenshots/error.png"
+```
+Returns base64 data ready for any image tool.
+
+### Media resolution control
+
+Reduce token usage by up to 75% whilst maintaining quality:
+
+| Level | Tokens | Savings | Best for |
+|-------|--------|---------|----------|
+| `MEDIA_RESOLUTION_LOW` | 280 | 75% | Simple tasks, bulk operations |
+| `MEDIA_RESOLUTION_MEDIUM` | 560 | 50% | PDFs/documents (OCR saturates here) |
+| `MEDIA_RESOLUTION_HIGH` | 1120 | default | Detailed analysis |
+| `MEDIA_RESOLUTION_ULTRA_HIGH` | 2000+ | per-image only | Maximum detail |
+
+For PDF OCR, MEDIUM gives identical text extraction quality to HIGH at half the tokens. Set `global_media_resolution` to apply to all images, or override per-image with `mediaResolution`.
+
+### Landing page generation
+
+```
+Use gemini:generate_landing_page with:
+  brief="A SaaS tool that helps developers monitor API latency"
+  companyName="PingWatch"
+  primaryColour="#6366F1"
+  style="startup"
+  sections=["hero", "features", "pricing", "cta"]
+```
+
+Returns a self-contained HTML file -- inline CSS and vanilla JS, no external dependencies. Styles: `minimal`, `bold`, `corporate`, `startup`.
+
+### Professional chart design systems
+
+The `gemini_prompt_assistant` tool includes 9 professional chart design systems:
+
+| System | Inspiration | Best for |
+|--------|------------|----------|
+| **storytelling** | Cole Nussbaumer Knaflic | Executive presentations -- everything muted except one bold highlight |
+| **financial** | Financial Times | Editorial journalism -- FT Pink background, serif titles |
+| **terminal** | Bloomberg / Fintech | High-density dark mode with electric neon |
+| **modernist** | W.E.B. Du Bois | Bold geometric blocks, stark contrasts |
+| **professional** | IBM Carbon / Tailwind | Enterprise dashboards |
+| **editorial** | FiveThirtyEight / Economist | Data journalism |
+| **scientific** | Nature / Science | Academic rigour |
+| **minimal** | Edward Tufte | Maximum data-ink ratio |
+| **dark** | Observable | Modern dark mode |
+
+```
+Use gemini:gemini_prompt_assistant with:
+  request_type="template"
+  use_case="product"
+  desired_outcome="Generate a professional product comparison chart"
+```
+
+### Help system
+
+```
+Use gemini:gemini_help with topic="overview"
+```
+
+Documentation for all features without leaving Claude. Topics: `overview`, `image_generation`, `image_editing`, `image_analysis`, `chat`, `deep_research`, `grounding`, `media_resolution`, `models`, `all`.
+
+---
+
+## Image output and storage
+
+**Default behaviour:** Images return as inline base64 previews (quality 100, 1024px) rendered directly in Claude.
+
+**Persistent storage:** Set `GEMINI_IMAGE_OUTPUT_DIR` to auto-save all generated images:
+
+```json
+"env": {
+  "GEMINI_API_KEY": "your-api-key-here",
+  "GEMINI_IMAGE_OUTPUT_DIR": "C:/Users/username/Pictures/gemini-output"
+}
+```
+
+Every image saves with a timestamp filename. The tool returns both the inline preview and the file path.
+
+**Per-call override:** Pass `outputPath` on any generation tool to save to a specific location.
+
+The server uses a two-tier compression approach to handle the MCP protocol's ~1MB JSON-RPC limit whilst preserving full-resolution files on disk:
+
+| Tier | Quality | Max dimension | Purpose |
+|------|---------|---------------|---------|
+| **Full-res** | Original | Original | Saved to disk |
+| **Viewer preview** | 100 | 1024px | MCP App inline preview (~400KB) |
+
+Gemini returns 2-5MB images. The full image is saved to disk immediately, and a compressed preview is created for the MCP App viewer.
+
+---
+
+## Configuration reference
+
+| Variable | Required | Default | Description |
+|----------|----------|---------|-------------|
+| `GEMINI_API_KEY` | Yes | -- | Google AI API key from [AI Studio](https://aistudio.google.com/apikey) |
+| `GEMINI_DEFAULT_MODEL` | No | `gemini-3.1-pro-preview` | Default model for `gemini_chat` and `analyze_image` |
+| `GEMINI_DEFAULT_GROUNDING` | No | `true` | Enable Google Search grounding by default |
+| `GEMINI_IMAGE_OUTPUT_DIR` | No | -- | Auto-save directory for generated images |
+| `GEMINI_ALLOW_EXPERIMENTAL` | No | `false` | Include experimental/preview models in auto-discovery |
+| `GEMINI_MCP_LOG_FILE` | No | `false` | Write logs to `~/.gemini-mcp/logs/` |
+| `DEBUG_MCP` | No | `false` | Log to stderr for debugging tool calls |
+
+---
+
+## Tools reference
+
+| Tool | Description |
+|------|-------------|
+| `gemini_chat` | Chat with Gemini 3.1 Pro. Google Search grounding on by default. Supports `thinking_level` for Gemini 3 |
+| `gemini_deep_research` | Multi-step iterative research with Google Search. Synthesises comprehensive reports |
+| `gemini_list_models` | Lists available models from the API |
+| `gemini_help` | Documentation for all features without leaving Claude |
+| `gemini_prompt_assistant` | Expert guidance for image generation with 9 chart design systems |
+| `generate_image` | Image generation with search grounding and thought signatures for conversational editing |
+| `edit_image` | Edit images with natural-language instructions. Supports multi-turn continuity |
+| `describe_image` | Fast image descriptions using Gemini 3 Flash |
+| `analyze_image` | Structured extraction and analysis using Gemini 3.1 Pro |
+| `load_image_from_path` | Read a local image file and return base64 for any image tool |
+| `generate_video` | Video generation with Veo 3.1 -- 4-8 seconds at up to 4K with native audio |
+| `generate_svg` | Production-ready SVG graphics for diagrams, illustrations, and data visualisations |
+| `generate_landing_page` | Self-contained HTML landing pages with inline CSS/JS |
+
+---
+
+## Model reference
+
+| Model | Used by | Notes |
+|-------|---------|-------|
+| `gemini-3.1-pro-preview` | `gemini_chat`, `analyze_image` | Default. Advanced reasoning |
+| `gemini-3-pro-image-preview` | `generate_image`, `edit_image` | Nano Banana Pro -- highest quality generation |
+| `gemini-2.5-flash-image` | `generate_image` (optional) | Faster generation, higher volume |
+| `gemini-3-flash-preview` | `describe_image` | Fast general descriptions |
+| `veo-3.1-generate-preview` | `generate_video` | Veo 3.1 -- 4K video with native audio |
+
+**Gemini 3 notes:** Temperature is forced to 1.0 on Gemini 3 models (Google's requirement -- lower values cause looping). Thought signatures are captured automatically for conversational image editing. Thinking level only applies to `gemini_chat`.
+
+---
+
+## Requirements
+
+- Node.js 18+
+- A Gemini API key from [Google AI Studio](https://aistudio.google.com/apikey)
+- ffmpeg (optional, for video thumbnail extraction)
+
+## Licence
+
+Apache-2.0