MCP Servers for Google Cloud Genmedia APIs

This repository provides Model Context Protocol (MCP) servers that enable AI agents and applications to easily integrate and leverage the full breadth and depth of Google Cloud's powerful generative media APIs (Gemini Image, Gemini TTS, Veo, Chirp, Lyria) and advanced audio/video compositing capabilities (AVTool).

Each server can be enabled and run separately, allowing flexibility for environments that don't require all capabilities.

Generative Media & Compositing Capabilities

Nano Banana: Gemini 3.1 Flash Image & Gemini 3 Pro Image - for image generation and editing
Veo 3 & 3.1 - for video creation
Gemini TTS & Chirp 3 HD - for speech synthesis
Lyria - for music generation
AVTool - for audio/video compositing and manipulation

🛠️ Agent Skills

We provide a set of expert Agent Skills that provide AI agents (like Gemini CLI or Antigravity) with deep domain knowledge on how to use these MCP tools effectively for complex workflows.

genmedia-producer: Orchestrates multi-step workflows like podcast creation and storyboarding.
genmedia-video-editor: Expertise in FFmpeg composition, image overlays, and GIF generation.
genmedia-audio-engineer: Specialist in high-fidelity TTS synthesis and multi-track mixing.
genmedia-image-artist: Expert in visual generation, prompt optimization, and collaborative refinement.
genmedia-voice-director: Expert in casting, directing, and generating expressive text-to-speech using Gemini TTS.
story-generator: Expert in generating full multi-scene multimedia storybooks (image, video, voice, and music) using direct, native MCP tool calls and parallel subagent orchestration, and packaging them into premium Material 3 interactive making-of reports.
install-mcp-genmedia: Installs the MCP servers from pre-compiled release binaries and registers them with your agent — no clone or Go toolchain required.
build-mcp-genmedia: Builds the MCP servers from source and wires them into your agent config, for environments where the pre-compiled binaries cannot be used.

See the Agent Skills directory for more information on how to link or install these skills.

Installation

Install MCP Servers: For detailed installation instructions, including an easy-to-use installer script, please refer to the Go Implementations README.

Easy Installation (Pre-compiled Binaries)

For the fastest setup without needing the Go toolchain installed, you can use our online installer script. This script automatically detects your operating system and architecture, downloads the latest pre-compiled binaries from GitHub Releases, and places them in your ~/.local/bin directory.

Run the following command in your terminal:

curl -sL https://raw.githubusercontent.com/GoogleCloudPlatform/genmedia-creative-studio/main/experiments/mcp-genmedia/mcp-genmedia-go/install-online.sh | bash

Note: Ensure ~/.local/bin is added to your system PATH.

Running the Servers

The MCP servers can be run using different transport protocols. The default is stdio.

To start a server in Streamable HTTP mode, use the --transport http flag:

mcp-nanobanana-go --transport http

Configuration

The servers are configured primarily through environment variables. Key variables include:

PROJECT_ID: Your Google Cloud project ID.
LOCATION: The Google Cloud region for the APIs (e.g., us-central1).
PORT: The port for the HTTP server (e.g., 8080).
GENMEDIA_BUCKET: The Google Cloud Storage bucket for media assets.

Available MCP Servers and Capabilities

Gemini Image Generate and edit images from text prompts.
Imagen: Generate images from text prompts (deprecated).
Gemini TTS Synthesize high-quality audio from text.
Gemini Transcribe: Transcribe pre-recorded audio files to text.
Veo: Create videos from text or images.
Gemini Omni: Create videos with optional embedded audio from text, optionally conditioned on input images or videos.
Lyria: Generate music from text prompts.
Chirp 3 HD: Synthesize high-quality audio from text.
AVTool: Perform audio/video compositing and manipulation (e.g., combining, concatenating, format conversion).

For a detailed list of tools provided by each server, refer to the Go Implementations README.

Authentication

The servers use Google's Application Default Credentials (ADC). Ensure you have authenticated by one of the following methods:

Set up ADC: gcloud auth application-default login
Set the GOOGLE_APPLICATION_CREDENTIALS environment variable to the path of your service account key file.

You may also need to grant your user or service account access to the Google Cloud Storage bucket:

gcloud storage buckets add-iam-policy-binding gs://BUCKET_NAME \
  --member=user:user@email.com \
  --role=roles/storage.objectUser

Client Configurations

The MCP servers can be used with various clients and hosts. A sample MCP configuration JSON can be found at genmedia-config.json.

This repository provides AI application samples for:

Development and Contribution

For those interested in extending the existing servers or creating new ones, the mcp-genmedia-go directory contains a more detailed README.md with information on the architecture and development process. Please refer to the mcp-genmedia-go/README.md for more information.

Contributions are welcome — see CONTRIBUTING.md for the contributor guide and CLA requirements.

License

Apache 2.0

Disclaimer

This is not an officially supported Google product.

MCP Servers for Google Cloud Genmedia APIs