Hello Apigee Community! 
Are you ready to scale your AI initiatives while maintaining absolute security and fine-grained access control? Moving agents into enterprise production requires a central “Control Tower” that can govern how models interact with your core systems.
In our latest TechTalk, Mathilde Leibenguth and Tyler Ayers from the Google Cloud team break down how Apigee acts as the ultimate access, proxy, and cataloging layer for Model Context Protocol (MCP) endpoints. They demonstrate how to safeguard your data warehouse using brand-new authorization techniques without complicating your cloud IAM roles. 
Watch the recording
Access the presentation here
Building the Control Tower Architecture 
To securely manage an agentic ecosystem, an enterprise needs three core capabilities at the gateway layer:
-
The Access Layer: Interacting natively with JSON-RPC 2.0 standards to unpack agent requests.
-
The Proxy Layer: Routing traffic dynamically to Google Cloud services like BigQuery, or external model endpoints.
-
The Cataloging Layer: Leveraging API Hub to automatically discover, version, and register filtered tool definitions for your AI platforms.
Key Takeaways from the Session 
During this highly interactive session, Mathilde and Tyler introduced the new Parse Payload policy now rolling out to Apigee. This powerful policy performs the heavy lifting by unpacking JSON-RPC 2.0 structures right at the gateway, allowing you to map explicit tool calls to standard API products and quotas seamlessly.
Tyler showcased a live demo using the Google Cloud BigQuery MCP server. He walked through how a single Parse Payload configuration can strip away dangerous write capabilities (like appending data to a data warehouse), leaving a perfectly safe, read-only tool specification for your agents. You will also see how to apply strict rate limits and quotas to specific tool calls, protecting your backend infrastructure from unexpected loops and scaling costs. Finally, the team discussed how API Hub automatically visualizes these filtered tool definitions down to the input and output schemas, building the perfect foundation for future AI documentation portals.
Next Steps 
Continue the Conversation: Are you experimenting with the new Parse Payload policy yet? Share how you plan to govern your custom MCP servers in the comments below!
Share Feedback: Help us shape future developer sessions by casting your vote and sharing your thoughts here → Feedback: Apigee Community TechTalks
Explore Further: Contact our sales team to map out your agentic proxy and gateway strategy → Contact Google Cloud Sales
Stay Tuned → Google Cloud Apigee Community TechTalks