External agents
External agents are AI agents you build and run outside Port. They fall into two common patterns:
- Code-first agents — built with SDKs in your own code, repos, or CI (for example OpenAI Agents SDK, Anthropic Claude Agent SDK, Claude Code, or agents running in GitHub Actions).
- Cloud-managed agents — hosted and operated on a vendor control plane (for example AWS Bedrock / AgentCore, Azure AI Foundry, or Cursor Cloud Agents).
Port connects to these agents so you can see which ones exist in your organization, what they can access, and trigger them from your workflows.
When teams lack a single view of external agents
As adoption grows, agents on external platforms often spread faster than your governance model:
- Teams deploy code-first agents in repos and cloud-managed agents on Bedrock, Foundry, Claude, and other platforms independently.
- Platform and security leads cannot answer basic questions: which agents are running, what data they can reach, or whether two teams built the same agent twice.
- Without inventory, you cannot assign ownership, enforce approval flows, or audit what an agent did after the fact.
What Port gives you
| Capability | What it means for you |
|---|---|
| Context Lake discovery | Agents, environments, sessions, skills, and vaults from connected platforms appear as entities in your Context Lake. |
| Workflow operations | Start sessions, send messages, or create agents through Port workflows instead of switching to each vendor console. |
| Data access control | Agents respect Context Lake permissions — access to data is scoped to what their role allows. |
| Cross-platform registry | Use an agent registry to track code-first and cloud-managed deployments in one view. |
Get started by platform
| Agent type | Examples | Documentation |
|---|---|---|
| Code-first | OpenAI Agents SDK, Anthropic Claude Agent SDK, Claude Code, GitHub Actions agents | Agent registry and Port MCP server for governed Context Lake access |
| Cloud-managed — AWS / Azure | AWS Bedrock / AgentCore, Azure AI Foundry | Agent registry and Build an AI Agent Registry guide |
| Cloud-managed — Cursor | Cursor Cloud Agents | Port MCP server for IDE and agent access to your Context Lake |
Related capabilities
- AI registry — govern skills, prompts, and MCP servers that agents use when they connect to Port.
- Port MCP server — connect IDEs and external tools to your Context Lake.