Agent registry
An agent registry is a governed inventory of external AI agents — both code-first agents built with SDKs and cloud-managed agents on platforms without a native Port connector — stored as entities in your Context Lake. Each registry entity tracks platform, status, model, region, and ownership so your organization can discover what exists instead of rebuilding agents in every squad.
Port's built-in _ai_agent blueprint can represent agents in your Context Lake. A dedicated registry blueprint (for example aiAgent) adds the metadata platform teams need for governance: approval status, owning team, deployment region, and which cloud platform hosts the agent.
What it solves
| Challenge | How a registry helps |
|---|---|
| No cross-platform visibility | View Bedrock, Azure AI Foundry, and other cloud agents in one place in your Context Lake instead of checking each vendor console. |
| Duplicate agents | Developers can search the registry before building a new agent and reuse what already exists. |
| Governance and compliance | Track which models, regions, and statuses each agent uses for audit and policy reviews. |
When to use a registry vs a native connector
| Approach | Best for |
|---|---|
| Native connector | Platforms with a dedicated Port integration that syncs agents, sessions, and related resources automatically. |
| Custom agent registry | Multi-cloud visibility when you ingest agents from Bedrock, Azure AI Foundry, or other APIs through a custom Ocean integration. |
Set up your registry
For blueprint JSON, integration mapping, and step-by-step setup for AWS Bedrock and Azure AI Foundry, see the guide Build an AI Agent Registry.
Where to go next
- External agents overview — how Port connects to agents hosted outside Port.
- Skills registry — govern reusable skills alongside your agent inventory.
- MCP registry — govern which MCP servers teams may use.