# Gemini MCP

> Delegates text tasks to Google Gemini through MCP (gemini/gemini-reply). Use for architecture disputes, design critique, second opinions. Don't use for media — use gemini-media for images/video.

- Skill: `neuron-one/gemini-mcp` (Agent Skill)
- Install (CLI): `npx skillmds@latest add neuron-one/gemini-mcp`
- Raw SKILL.md: https://api.skillmd.com/api/skills/neuron-one/gemini-mcp/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: neuron-one (https://skillmd.com/u/neuron-one)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/neuron-one/gemini-mcp

---


# Gemini MCP Integration

## Two Gemini Servers

### `gemini` (Text — Delegator Bridge)
- Tools: `mcp__gemini__gemini`, `mcp__gemini__gemini-reply`
- Use for: architecture disputes, code review, design critique, text consultations
- Powered by: Gemini CLI → Gemini 2.5 Flash/Pro
- Supports: multi-turn (threadId), developer-instructions, sandbox modes

### `gemini-media` (Media — @fre4x/gemini)
- Tools: `mcp__gemini-media__analyze_media`, `mcp__gemini-media__generate_image`, `mcp__gemini-media__generate_video`
- Use for: image analysis, image generation (Imagen), video generation (Veo)
- Powered by: Gemini API directly

## When to Use Gemini (Text)
- As independent expert in /dispute (full text architecture/design reviews)
- Design and visual critique (describe, don't just send image)
- Second opinion on Claude's decisions
- Tasks where Google ecosystem knowledge helps

## When to Use Gemini-Media
- Analyze screenshots, diagrams, photos
- Generate images (Imagen)
- Generate videos (Veo)

## Rules
- Include full context (Gemini doesn't see our conversation)
- Specify output format expected
- Compare with Claude's own analysis
- Note: data sent to Google servers