A2A interoperability

How do I connect AI agents from different vendors so they can delegate tasks to each other?

Connect compatible agents through an agent-to-agent boundary that uses agent identities, cards, credentials, and routed messages instead of exposing private runtime internals. Hermes Hub provides shipped A2A edge support for this model, while each connected agent and delegation policy still needs to be evaluated by the operating team.

What an interoperable delegation path needs

Protocol support is only the connection layer. Define identity, capability, authorization, routing, and evidence around it.

Agent cards

Expose the agent's supported edge identity and connection information at the boundary.

Credentials

Issue and manage credentials for the intended peer relationship.

Routed messages

Route task requests through a controlled edge instead of giving peers direct runtime access.

Policy and evidence

Set delegation boundaries and retain the operational record needed to review the exchange.

Build the delegation architecture

A2A edge

Connect compatible outside agents through controlled cards, credentials, peers, and messages.

Learn more

Cross-vendor agent delegation FAQ

Is A2A shipped in Hermes Hub?
Yes. Hermes Hub’s shipped scope includes A2A edge support with agent cards, credentials, peers, and routed messages.
Does A2A expose the private agent runtime?
The intended boundary is controlled edge access rather than direct access to tenant runtime internals.
Is A2UI the same as A2A?
No. A2A is shipped. Full A2UI remains roadmap; only a strict limited actions subset is currently described as existing.