Integrations¶
Agent-manifest is framework-agnostic - it is a signing and verification layer, not an agent runtime. These guides show how to attach it to the most common Python agent frameworks.
| Integration | What it covers |
|---|---|
| LangChain | Manifest for a LangChain agent; tool wrapper that verifies caller manifests |
| OpenAI Agents SDK | Manifest per agent; manifest handoff verification during agent handoffs |
| AutoGen and CrewAI | Per-agent manifests in AutoGen conversations and CrewAI crews |
| AGT (Agent Governance Toolkit) | Using agent-manifest as the identity layer feeding AGT policy and trust scores |
Common pattern¶
Every integration follows the same three steps:
- Issue - create and sign a manifest that identifies the agent and binds its artifacts
- Attach - pass the
manifest_idto the relying party (header, metadata, context field) - Verify - the relying party calls
verify_manifest()before acting on the agent's output
The framework-specific pages show where exactly to hook each step.