Apigee MCP support | Google Cloud Blog

time tools/list or tools/call When a request is made to an MCP endpoint, Apigee uses the operations described in the OpenAPI specification as the MCP tool list. You can then bundle your MCP proxy. API productsapply granular quotas, identity, and access policies to ensure that only authorized MCP clients, agents, and developers can list and invoke these tools.

This process allows you, for example, to designate certain API operations as MCP tools. You can then specify that the Payments service's MCP tool can only be accessed by designated agents with known client IDs and a legitimate need to use the tool.

You can then monitor your MCP tool usage using Apigee Analytics. And with the recent release of Apigee, API insightsnew in the Apigee API hub catalog.[インサイト]You can also use the tabs to view traffic and performance metrics for your MCP endpoints.

Benefits of Apigee's approach to MCP support

The primary goal of MCP support in Apigee is to enable you to secure, manage, and monitor your use of MCP tools using the same Apigee policies and workflows that you are already familiar with.

What this means for you:

No additional operational burden: There is no need to set up and manage an MCP server for each API. Just deploy the MCP proxy and Apigee takes care of the rest. Apigee fully manages MCP servers, transcoding, and protocol processing.
Tool observability and governance: You can also use Apigee's built-in identity, authorization, and security policies to secure and manage your MCP endpoints and tools. You can also use Apigee Analytics to monitor tool usage by MCP clients.
Comprehensive tool security: Apigee helps ensure that all agent interactions are secure. For example, you can use Cloud data loss prevention Classify and protect sensitive data passed from tools, model armor To prevent rapid injection or jailbreak attempts. You can ensure that your agents and users have the appropriate IAM permissions to call MCP tools and view and fully debug the entire end-to-end flow of agent interactions. can also be used Apigee's advanced API security To keep your tools safe.
Centralized tool catalog: When you deploy an MCP proxy, Apigee automatically registers the MCP endpoint with your specification in the Apigee API hub. This results in searchablecreate a centralized tool catalog and promote tool reuse.

Using Apigee MCP tools with the agent framework

Apigee's MCP support is designed to maximize compatibility. Protected Apigee MCP endpoints can be used with agents built using various frameworks, such as ADK., Langgraf, Other popular solutions across the AI ecosystem.

However, the developer chose agent development kit (ADK) There are unique and logical advantages when developing agents within the Google ecosystem.

ADK is a flexible, modular framework for developing and deploying AI agents. ADK is optimized for Gemini and the Google ecosystem, but is model and deployment agnostic and built with compatibility with other frameworks in mind. The ADK is designed to make agent development feel like software development, allowing developers to easily create, deploy, and adjust agent architectures ranging from simple tasks to complex workflows.

ADK includes toolset Apigee and Application integrationThis allows developers who build custom agents using the ADK to easily connect those agents to MCP endpoints and tools managed and secured by Apigee. You can also use ApigeeLLM wrapper for ADK Expose LLM endpoints through Apigee proxies to integrate governance into agent workflows. (Note: ApigeeLLM The wrapper is currently designed for use with Vertex AI and Gemini APIs. Google AI StudioHowever, we plan to support other models and interfaces as well).

Google Cloud also provides services for deploying custom agents. can be used Vertex AI agent engine Deploy the agent and run it across your organization using: gemini enterprise.