Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/sandraschi/google-ai-mcp/agents-mdgit clone --depth 1 https://github.com/sandraschi/google-ai-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/sandraschi/google-ai-mcp/agents-md)<a href="https://agentmods.dev/instructions/sandraschi/google-ai-mcp/agents-md"><img src="https://agentmods.dev/badge/instructions/sandraschi/google-ai-mcp/agents-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00781 | $0.00781 |
| Opus 5 | $0.00391 | $0.00391 |
| Sonnet 5 | $0.00156 | $0.00156 |
| Haiku 4.5 | $0.00078 | $0.00078 |
Grade A, and why
google-ai-mcp AGENTS.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 89 lines — stays where its author put it; the contents beside it link to each section on GitHub.
google-ai-mcp — Agent Guide
FastMCP 3.2 MCP server for Google AI services — Gemini Chat, Nano Banana image gen, Veo video, Lyria music, Omni, one-shot TTS, embeddings, and LeWM world-model bridge.
Speech boundary: streaming Live/STT lives in speech-mcp; this repo keeps thin TTS only.
Entry Points
uv run google-ai-mcp→google_ai_mcp.server:main.\start.ps1— backend + frontendjust mcpb-pack— Claude Desktop.mcpbbundle
Quick Ref
uv run python -m google_ai_mcp.server
uv run pytest tests/ -q
just e2e
uv run ruff check src/ tests/
Ports
| Service | Port |
|---|---|
| Backend (FastAPI + FastMCP HTTP) | 11014 |
| Frontend (Vite React) | 11015 |
Architecture
src/google_ai_mcp/
server.py — FastMCP 3.2 server, MCP tools
web.py — FastAPI REST endpoints
omni_client.py — Gemini Omni multimodal
chat_client.py — Gemini/Gemma chat
imagen_client.py — Nano Banana image gen
veo_client.py — Veo 3.1 video gen
lyria_client.py — Lyria 3 music gen
speech_client.py — One-shot TTS (not Live)
embeddings_client.py — Text embeddings
clients_registry.py — LRU-cached client factory
webapp/ — React MUI SPA (Vite)
native/ — Tauri 2.0 + PyInstaller sidecar
manifest.json — MCPB Claude Desktop bundle
MCP Tools
| Tool | Operations |
|---|---|
google_ai_chat |
chat, list_models |
google_ai_image |
generate, list_models |
google_ai_video |
generate, list_models |
google_ai_omni |
generate, list_models |
google_ai_music |
generate, list_models |
google_ai_speech |
tts, list_voices, list_models |
google_ai_embeddings |
embed, list_models |
google_ai_world |
health, train_prepare, infer_prepare |
google_ai_status |
Health check all services |
show_google_ai_status_card |
Prefab UI card |
Standards
- FastMCP 3.2+ portmanteau tool pattern — tools use
operationenum param - Responses: structured dicts with
success,message, domain-specific fields - Dual transport: stdio (Claude Desktop) + HTTP (
MCP_TRANSPORT=http) - See mcp-central-docs for fleet-wide coding standards
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 5d ago First seen · 89 lines · 781 tokens per session scan A e732ca135368
google-ai-mcp AGENTS.md is an instructions file published in the GitHub repository sandraschi/google-ai-mcp (2 stars, last pushed 5d ago), licensed MIT. It adds 781 tokens to every session, about $0.0039 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
agent-starter-pack GEMINI.md
Instructions for GoogleCloudPlatform/agent-starter-pack, covering agent starter pack - ai coding agent guide, core principles for ai agents, project architecture overview, 4-layer template system and key directory structure.
civitai-mcp AGENTS.md
AGENTS.md instructions for sandraschi/civitai-mcp, covering civitai-mcp — agent guide, standards, key files and quick ref.
civitai-mcp CLAUDE.md
Claude Code instructions for sandraschi/civitai-mcp, covering civitai-mcp — claude / agent context, do, don't and commands.
OmniMCP CLAUDE.md
Claude Code instructions for OpenAdaptAI/OmniMCP, covering claude.md - omnimcp implementation guide, core architecture, implementation approach, 1. start with visualstate and 2. implement core mcp server.
doc-adk-training GEMINI.md
Gemini CLI instructions for mauripsale/doc-adk-training, covering adk docs migration project guidelines, core mandates and how to run the student evaluator.
quartier CLAUDE.md
Instructions for EnriqueLop/quartier, covering quartier, testing, claude-specific and skill routing.