Gemini Agent Platform: Build Agents Faster with Remote MCP
Quick answer
Google Cloud's Gemini Enterprise Agent Platform now offers a fully-managed remote MCP server, letting external AI agents securely access cloud resources with ease.
Google Cloud’s Gemini Enterprise Agent Platform just got a serious upgrade: a fully-managed, remote MCP server that lets your external AI agents securely tap into your Google Cloud resources. Think of it as a bridge between your favorite IDE and your cloud architecture—no more wrestling with custom integration code.
Why This Matters
Developers want speed; IT needs governance. This MCP server gives you both. Your agents (built in Antigravity CLI, Claude Code, or any MCP-compliant framework) can now call models from Model Garden, pull prompt templates, or manage Notebooks—all without leaving your IDE. It’s like having a capybara-friendly channel through the swamp: smooth, secure, and surprisingly fast.
Key Benefits
- Open standards, no lock-in: Your agents stay compliant with the open MCP spec. No proprietary chains here.
- Centralized discovery: Agent Registry acts as your team’s library for skills, tools, and AI capabilities.
- Security by default: Cloud IAM Deny policies ensure external frameworks only touch authorized resources.
Get Started in 3 Steps
- Enable the API: It’s automatically enabled when you turn on the Gemini Enterprise Agent Platform API in your Google Cloud project.
- Configure your client: Point your AI app to the remote server using the configuration docs.
- Use toolsets: Access endpoints like
/mcp/generate,/mcp/predict,/mcp/notebook, and more to start building immediately.
Available Toolsets
| Endpoint | Description | Tools |
|---|---|---|
| /mcp/generate | Generative AI tools | Core generation features |
| /mcp/predict | Prediction tools | Inference and raw prediction |
| /mcp/notebook | Colab enterprise notebook tools | Notebook runtime and execution management |
| /mcp/endpoints | Endpoint management tools | Lifecycle management for model endpoints |
| /mcp/models | Model registry tools | Model upload, registry, and deployment |
| /mcp/tuning | Model fine-tuning tools | Finetuning job management and tracking |
| /mcp/evaluation | Quality evaluation tools | Automated model quality and instance evaluation |
| /mcp/prompts | Prompt management tools | Prompt engineering and versioning workflows |
Ready to dive in? Check out the Agent Platform docs and start building agents that glide through your cloud swamp like a pro capybara.
For more on Google Cloud’s offerings, see our Google Cloud Review.
Original announcement published on Google Cloud.