Discovery
Agents can discover capability documents through a standard well-known path and through registries or catalogs.
Agent2Agent protocol
From readable websites to discoverable agent interfaces
A2A defines how agents discover each other, understand capabilities, negotiate interaction modes, and exchange tasks without exposing internal tools or private state. AgntBase uses this direction as a product map for business identity, discovery and trust.
Documentation
What A2A changes
A2A turns agent interoperability into a public web concern: discovery, identity, capabilities, auth and task exchange become inspectable objects.
Capability negotiation
Agent Cards describe skills, supported interfaces, input and output modes, versions and docs.
Task lifecycle
A2A is not just chat. It supports messages, stateful tasks, streaming, auth-required states and completion flows.
Documentation
How this maps to AgntBase
A2A is protocol infrastructure. AgntBase focuses on the business layer that agents need before interaction becomes useful.
Canonical profile
One owner-approved source of truth for the company, services, trust signals, limits and next actions.
Registry and discovery
A normalized index of profiles, files, agent cards and readiness signals that can be checked and compared.
Trust and governance
Provenance, freshness, noindex policy for registry pages, bot analytics and future signed Agent Cards.
Documentation
Official sources
We use current A2A v1 documentation as the direction of travel and avoid older pre-v1 discovery examples.
| Specification | A2A latest specification |
|---|---|
| v1.0 announcement | Announcing A2A v1.0 |
| A2A and MCP | Complementary, not competing |
| Protocol bindings | Custom protocol bindings |
| Roadmap | A2A roadmap |
| Protocol repository | a2aproject/A2A on GitHub |
| Important warning | Use /.well-known/agent-card.json. Older agent.json examples should not be treated as the current default. |