Apigee

Prev Next

The PlainID Apigee Authorizer enforces PlainID Policies on REST APIs and Model Context Protocol (MCP) servers exposed through Google Apigee, with minimal involvement from developers. Beyond enforcement, the Authorizer also discovers the AI agents and MCP tools that pass through Apigee, so the Platform keeps an up-to-date inventory of what agents exist and what tools they can reach. The Authorizer ships as a set of Apigee Shared Flows. Flow Callout policies attach the required Shared Flows to API proxies, and Apigee consults the PlainID Policy Decision Point (PDP) at request time.

The Authorizer provides five capabilities:

  • API authorization: for each REST call, the PDP decides whether to permit or deny the request. Apigee blocks denied calls before they reach the backend service.
  • MCP tool governance: for MCP servers, the Authorizer filters the tool list each identity receives and authorizes every tool call.
  • Agent discovery: the Authorizer identifies the AI agents that call API proxies and MCP servers through Apigee and registers them with PlainID.
  • MCP tool discovery: the Authorizer registers the full tool inventory of each MCP server with PlainID.
  • Insights: the Authorizer intercepts the transactions that pass through Apigee and reports them to the Platform.

Agent discovery, MCP tool discovery, and Insights report to an Apigee POP in the Integrations Workspace, which lists the agents, MCP servers, and tools the Platform discovers.

See Configuring the Apigee Authorizer and POP Setup for more information.

How It Works

REST API Requests

  1. An application calls an API proxy in Apigee. When Insights is enabled on the proxy, the Authorizer reports the request to PlainID.
  2. A Flow Callout in the proxy calls the PlainID Authorizer, a Shared Flow that can serve multiple proxies.
  3. The Authorizer requests an access decision from the PDP, which calculates it from the Policies configured in the Platform.
  4. Apigee enforces the decision: it denies the request, permits it as is, or permits it with PlainID enriching the access token with further authorization instructions.

MCP Requests

  1. An agent asks the MCP server for its tools (tools/list). When Insights is enabled, the Authorizer reports the request to PlainID.
  2. The Authorizer fetches the complete tool list from the MCP server and registers it with PlainID. It then asks PlainID which tools this identity may use, removes the rest, and answers the agent itself.
  3. The agent calls a tool (tools/call). Before the call reaches the MCP server, the Authorizer sends the tool name and arguments to the PDP.
  4. On Permit, Apigee forwards the call to the MCP server. On Deny, Apigee returns a JSON-RPC error to the agent, and the call never reaches the MCP server.

Dynamic Discovery

The Authorizer discovers agents and tools dynamically from live traffic, so you don't need to register them in the Platform manually.

PlainID discovers agents, NHIs, and tools dynamically from the traffic that passes through Apigee, so you don’t need to register them in the Platform manually.

  1. The Authorizer sends transaction data from the API proxies and MCP servers in Apigee to the PlainID data pipeline, which ingests all gateway traffic.
  2. When an agent calls tools/list, the Authorizer also fetches the complete tool inventory from the MCP server and registers it with PlainID, including tools the current identity isn’t permitted to use.
  3. PlainID discovers the agents, NHIs, and tools in that traffic and lists them in the Apigee POP in the Integrations Workspace. The Platform updates the inventory as new agents appear or an MCP server’s tool list changes.
  4. The discovered entities feed two areas of the Platform: agentic insights, which show how agents and tools are used, and Policy authoring, where you use the discovered agents and tools to define Policies.

See Configuring the Apigee Authorizer for installation and configuration steps.

Needs your review: I don't have the Kong, AgentCore, and APIM pages in front of me, so the Dynamic Discovery steps describe the behavior generically. Please check two details against those pages: how the Apigee Authorizer identifies an agent (which token claim or header it reads), and whether step 4 matches how discovered agents and tools become usable in Policies (for example, as Assets).

© 2026 PlainID LTD. All rights reserved.