A declarative specification for AI agent identity, authority, and operational boundaries — making agent identity, declared authority, and operational boundaries legible before execution.
Agent Manifest is an open specification that enables AI agents to declare their identity, purpose, capabilities, and operational boundaries in a structured, machine-readable format before any interaction begins.
A manifest is a pre-execution declaration. It is not a runtime observation, a behavioral log, or an enforcement mechanism. It is a structured commitment made by or on behalf of an agent, published at a known location, and readable by any party before the interaction starts.
- It is not a runtime monitor or behavioral inspector.
- It is not a compliance enforcement layer.
- It is not a guarantee of behavior — it is a declaration of intent and declared constraints.
- It does not observe, intercept, or audit live agent execution.
- It does not replace governance frameworks, policy engines, or audit systems.
A manifest makes declared identity and boundaries legible. Whether those declarations are honored is a separate layer of responsibility.
Declaration before interaction. An agent that declares itself before acting is structurally different from one that does not. The declaration creates a record, a reference point, and an implicit commitment.
Pre-execution legibility. Identity, purpose, capabilities, and boundaries must be readable before execution begins — not reconstructed from logs after the fact.
Structural accountability. Public declaration under a known identity creates accountability through visibility, not through enforcement.
Open and composable. The specification is open, versioned, and designed to be composed with governance, policy, and discovery layers built by others.
| Repository | Role |
|---|---|
| agent-manifest | Core specification and JSON schema |
| agent-manifest-registry | Governance model and discovery layer |
| agent-manifest-dataset | Public registry of declared manifests |
| agent-manifest-ambassador | Conversational manifest generator |
| agent-manifest-diplomat | Registration API endpoint (infrastructure) |
| boundary-handshake | Declaration-first agent interoperability |
| e-principle | Philosophical foundation: The Destination of Actions |
Schema
https://agent-manifest-spec.org/spec/v1.0/schema.json
Discovery endpoint
https://agent-manifest-spec.org/.well-known/agent-manifest-registry.json
Website agent-manifest-spec.org
Agent Manifest ecosystem artifacts are archived with DOI via Zenodo.
| Artifact | DOI |
|---|---|
| Agent Manifest specification | 10.5281/zenodo.18833956 |
| Boundary Handshake | 10.5281/zenodo.18894794 |
| ∈ Principle | 10.5281/zenodo.18965669 |
ORCID 0009-0008-7216-3032
Specification and documentation: CC BY 4.0 Infrastructure code: MIT
Individual repositories may carry their own license terms.