Agent-to-Agent (A2A)
Open protocol · Agent interoperability and the Agent Card SIGNET support: Architectural alignment — capability discovery
What it is
A2A is an open protocol for interoperability between AI agents across organisational boundaries. Its discovery model is the Agent Card: an agent publishes a description of itself and its capabilities at a well-known location (/.well-known/agent.json), so other agents can find and understand it without prior arrangement.
How SIGNET supports it
A2A sits in SIGNET's architectural alignment tier, and the alignment is explicit in the specification: the SyntheticAgent object carries an agentCard reference, and the AgentCapability model aligns to the A2A Agent Card, so an agent's declared capabilities in the canonical model correspond to what it advertises at its well-known location.
The two standards divide the problem cleanly. A2A answers how agents find and talk to each other; SIGNET answers what a commerce agent is, what it is authorised to do, and how its actions are held to account. An A2A-discoverable agent operating in a SIGNET network is a first-class Party with declared capabilities, bounded by a Mandate, producing accountable Decision records with cryptographic provenance. Discovery gets an agent to the table; the Agent Layer governs what happens once it is there.
This is the pattern across SIGNET's AI-protocol alignments: the emerging agent protocols are adopted for what they define, and the canonical model supplies the governance layer that none of them attempts.
Where it lives in the standard
- Agent Layer — SyntheticAgent (
agentCard), AgentCapability - Standards Mapping — the A2A alignment
- Schema source:
synthetic-agent.schema.json