Platform Discovery — Where AIControl's Gateway Plugs In

For each platform below, this is the specific, already-documented, customer-controllable insertion point your agent's outbound tool call can be pointed at.

Amazon Bedrock

Classic Agents route through your own Lambda action-group code — a code-level insertion point, not a config field. Point that Lambda's outbound call at AIControl's gateway the same way any other MCP-capable code would.

AgentCore Gateway agents point their MCP client at a Gateway endpoint URL — the same shape as pointing at AIControl directly. You choose which gateway sits in that position.

Azure AI Foundry (Microsoft Foundry)

An OpenAPI tool's servers.url field, or a Foundry Toolbox's MCP endpoint — both plain, documented configuration fields. Point either at AIControl's /mcp/{server_id}/ endpoint.

Vertex AI / Gemini Enterprise Agent Platform

Materially different from the other three platforms here. Google ships its own "Agent Gateway" as the platform's data plane; it calls out to an external authorization service via Service Extensions. On this platform, AIControl's integration shape is as the delegated authorization backend Agent Gateway calls — not as a front-door gateway replacing it. Don't expect "point your MCP client at AIControl" to apply here the way it does for Bedrock, Azure, or Copilot Studio.

Microsoft Copilot Studio

A custom connector's Host/Base URL field, or the newer REST API tool (preview at time of writing, not production-ready per Microsoft's own docs — confirm current status before relying on it). Point either at AIControl's /mcp/{server_id}/ endpoint.

---

Custom agent frameworks

See the Platform & Framework Guide for LangChain, LangGraph, CrewAI, Google ADK, and other MCP-client-capable frameworks — those are a different category from the four platforms above, each of which has its own platform-specific insertion point rather than a generic MCP client.