Open Specification

Agent Manifest

A minimal declaration layer for autonomous AI systems

https://doi.org/10.5281/zenodo.18833956 — CC BY 4.0

Concept

Agent Manifest allows autonomous systems to declare identity, purpose, capabilities, and operational boundaries before interaction. This declaration creates the surface on which validation, auditability, and governance become possible.

The specification declares. It does not execute, validate, score, enforce, or decide.

  • What this is

    The scope of the specification, and what falls outside it.

    /WHAT_THIS_IS.html
  • Why this exists

    The problem the declaration layer addresses.

    /WHY_THIS_EXISTS.html
  • Core principles

    The constraints that hold across the v1.x series.

    /CORE_PRINCIPLES.html
  • Terminology

    The governed vocabulary of the specification.

    /TERMINOLOGY.html
  • Design rationale

    Why each field is defined as it is, and why others are not.

    /DESIGN_RATIONALE.html
  • Common misconceptions

    Clarifications of recurring misunderstandings about scope.

    /docs/MISCONCEPTIONS.html

Specification

  • Specification v1.0

    The canonical declarative specification.

    /spec/v1.0/agent_manifest_v1.0.html
  • JSON Schema

    Normative schema for structural validation.

    /spec/v1.0/schema.json
  • Version index

    Release status, stability commitment, and an example manifest.

    /spec/v1.0/
  • Repository

    Source, issues, and revision history.

    github.com/agent-manifest/agent-manifest

Documentation

  • Documentation

    Reference and guides for the specification.

    /docs/
  • Reading a manifest from your own software

    For someone building software that could read a declaration: install the reading client, resolve a published manifest, and branch on a declared field. Four worked examples. Package @agent-manifest/client.

    /docs/consume/
  • Command-line interface (CLI)

    The Agent Manifest CLI — the command-line validator for v1.0 declarations. Package @agent-manifest/cli, command agent-manifest: npm install -g @agent-manifest/cli.

    /docs/cli/
  • How to register an agent

    Steps to publish a manifest to the registry.

    /docs/how-to-register.html
  • Architecture

    The three-layer declaration model.

    /docs/ARCHITECTURE.html
  • Operational architecture

    How declaration, enforcement, and execution relate.

    /docs/architecture/system-flow.html
  • Ecosystem map

    The repositories of the ecosystem and the role of each.

    /docs/ecosystem-map.html
  • Relation to adjacent specifications

    Where Agent Manifest sits beside A2A, MCP, agents.json, OASF, and Agent Spec.

    /docs/comparison.html
  • Examples

    Non-normative manifests covering the declared field surface.

    /examples/

Research

  • Academic Surface

    The ecosystem’s citable record of works.

    /works/
  • ∈ Principle

    Foundational statement on public authorship in the age of AI.

    /works/e-principle/
  • Declaration Layers

    Working paper on the evaluation of agent boundaries.

    /works/declaration-layers/
  • Foundations

    Non-normative doctrinal notes underlying the specification.

    /foundations/
  • Evaluations

    Worked applications of the declarative integrity checklist.

    /evaluations/
  • Papers

    Writing by the author on the declaration layer.

    /PAPERS.html
  • External citations

    Third-party works that cite Agent Manifest.

    /CITATIONS.html
  • Public comment — NIST NCCoE

    Comment submitted on the agent-security concept paper.

    /nist-comment/nist-comment-2026.html

Infrastructure

  • Ambassador

    Guided manifest authoring and submission.

    agent-manifest.github.io/agent-manifest-ambassador
  • Diplomat API

    Manifest registration endpoint.

    POST /api/register
  • Dataset

    Public corpus of registered manifests.

    github.com/agent-manifest/agent-manifest-dataset
  • Registry

    Governance and indexing of declarations.

    github.com/agent-manifest/agent-manifest-registry
  • Command-line interface (CLI) — repository

    Source of the Agent Manifest CLI. Documented at /docs/cli/.

    github.com/agent-manifest/agent-manifest-cli

Operational pipeline

Ambassador → Diplomat → Dataset → Registry → Discovery

Status

Specification v1.0
Published
Ambassador
Operational
Diplomat API
Operational
Dataset
Public
Registry index
Active
Discovery endpoint
Active
Command-line interface (CLI)
Published — @agent-manifest/cli 0.1.2

Discovery

Governance

  • Governance

    Stewardship, the public surface principle, and contributions.

    /GOVERNANCE.html
  • Versioning policy

    How the repository version and the specification version relate.

    /VERSIONING_POLICY.html
  • Stability

    What is frozen at v1.0 and what may still change.

    /STABILITY.html
  • Changelog

    Dated record of releases and changes.

    /CHANGELOG.html
  • Roadmap

    Stated direction, and what is explicitly not planned.

    /ROADMAP.html
  • Security policy

    How to report a vulnerability and what the project does not certify.

    /SECURITY.html
  • Security review context

    What a declared manifest contributes to a security review.

    /SECURITY_REVIEW_CONTEXT.html
  • Contact

    The institutional contact path, and who maintains the specification.

    /contact/
  • Privacy

    What the site collects — nothing — and which third parties serve a request.

    /privacy/
  • Contributing

    Which contributions are accepted, and which are out of scope.

    CONTRIBUTING.md
  • Code of conduct

    Adopted from the Contributor Covenant v2.0.

    CODE_OF_CONDUCT.md