ADL vs A2A (Agent2Agent)
Short answer: ADL and A2A solve different problems and are designed to work together. A2A (Agent2Agent) is a protocol for how agents communicate, advertising their capabilities through an Agent Card that a client can retrieve to learn how to interact. ADL (Agent Definition Language) describes what an agent is — its identity, permissions, security, lifecycle, and compliance — in one auditable document. ADL is a superset of the identity information an Agent Card carries, and an ADL document can generate an A2A Agent Card. You do not choose one over the other; you use ADL to govern the agent and A2A to connect it.
What is A2A?
A2A (Agent2Agent), introduced by Google and now stewarded under the Linux Foundation, is an open protocol for agent-to-agent communication. An agent publishes an Agent Card — a JSON document advertising its name, description, skills, and endpoints — typically served at a well-known URL. A client that already knows the agent's address, or finds it through a separate registry or catalog, retrieves the card to learn the agent's capabilities and how to invoke them. A2A standardizes this capability discovery and the interaction that follows; it does not automatically discover unknown agents across a network — that requires a registry or directory layered on top.
What is ADL?
ADL is an open standard for defining an AI agent in a single machine-readable document: identity (including cryptographic identity), capabilities and tools, permission boundaries, security requirements, data classification, lifecycle, and compliance posture. It is the agent's "passport" — the authoritative, auditable record of who the agent is and what it is authorized to do.
Key differences
| A2A (Agent Cards) | ADL | |
|---|---|---|
| Primary purpose | How agents communicate and advertise capabilities | What an agent is and may do |
| Scope | Capabilities and endpoints for interaction | Identity, permissions, security, lifecycle, compliance |
| Discovery | A known agent's Agent Card (per agent) | Per-domain directory at /.well-known/adl-agents |
| Identity | Partial (name, endpoint) | Cryptographic identity (DID) + attestation |
| Permissions | Not a permissions model | Deny-by-default, auditable boundaries |
| Governance / compliance | Out of scope | NIST 800-53, SOC 2, ISO 27001, EU AI Act mappings |
| Lifecycle | Partial | Status, versioning, sunset dates, successors |
| Relationship | Consumes identity | Generates A2A Agent Cards |
Discovery
Both standards use well-known URLs, but at different granularities. An A2A Agent Card describes a single agent, so a client needs that agent's address — or a separate registry — to find it. ADL additionally defines a per-domain discovery document at https://{domain}/.well-known/adl-agents (§6.4): a directory of the agents a domain publishes, each with a short description and keywords so a client can triage which ones to retrieve in full. It still requires knowing the domain — it is not global, network-wide discovery of unknown agents — and what a discovering agent may then engage is bounded by its permissions.delegation envelope and admitted through the Trust Protocol, so discovery widens reach without widening authority.
How ADL and A2A work together
ADL sits above A2A. You author one ADL document as the source of truth for the agent, then generate an A2A Agent Card from it for capability discovery and communication. Per the ADL specification §15.1, an implementation maps ADL fields to an Agent Card: name, description, version, tools → skills, cryptographic_identity.did → id, and security.authentication → authentication.
The result: the Agent Card an agent advertises over A2A is provably derived from the same governed, auditable ADL passport that defines its permissions and compliance — so communication and governance stay in sync.
When to use which
- Use A2A when you need agents to advertise their capabilities and call each other.
- Use ADL when you need a portable, auditable definition of an agent's identity, permissions, and compliance — and to generate the Agent Card from it.
- Use both together for interoperable agents that are also governed and audit-ready.
FAQ
Is ADL a replacement for A2A?
No. They operate at different layers. A2A handles agent-to-agent communication; ADL handles agent identity and governance. ADL complements A2A and can generate A2A Agent Cards.
Can I use ADL and A2A together?
Yes — that's the intended pattern. Define the agent in ADL, then generate its A2A Agent Card from the ADL document so capability discovery and governance share one source of truth.
Does ADL replace the Agent Card?
No. The Agent Card remains the A2A discovery artifact; ADL is the broader passport it can be generated from.