pe_gcp_agent

An MCP server giving AI agents direct access to Google Cloud — automate GCP resource management, monitoring, and operations with secure service-account auth and least-privilege controls.

Comment

Inspiration

We wanted to give AI agents a safe, structured way to operate on Google Cloud Platform — automating routine resource management and monitoring tasks without exposing raw cloud credentials to the model.

What it does

pe_gcp_agent is an MCP (Model Context Protocol) server that exposes Google Cloud Platform capabilities as tools for AI agents. It enables automated resource management and monitoring across GCP services through a standardized agent-friendly interface.

How we built it

  • Python 3.8+ project structured as an MCP server (main.py entry point)
  • Authenticates to GCP using a service account via the GOOGLE_APPLICATION_CREDENTIALS environment variable
  • Tools surfaced through the MCP protocol so any compatible agent (e.g., Gemini) can call them
  • Unit tests under tests/ using pytest

Challenges we ran into

Designing the right tool surface for GCP without overexposing the API, and enforcing least-privilege scopes on the service account.

What we learned

How to wrap large cloud APIs into clean, agent-callable MCP tools, and how to keep credentials secure when bridging LLMs to production cloud services.

What's next

Expanding coverage to more GCP services, adding richer monitoring tools, and improving observability of agent-driven cloud actions.

Built With

Share this project:

Updates