Dual identity validation
Agent token and user context token, both required.
Zero-Trust Agent Security
Every agent action traces back to a human identity.
Dual identity validation, purpose-scoped token exchange, and fail-closed enforcement on every agent-to-agent request.
- Human userOriginal identity
- Agent AAgent token + user context token
- FabricValidates both, exchanges a scoped token (RFC 8693)
- Agent BReceives a purpose-scoped token
- 2
- identities validated on every A2A request
- Per hop
- purpose-scoped token exchange
- Fail-closed
- unverifiable policy means no traffic
The problem
As agents start calling other agents, "who authorized this" gets harder to answer. A compromised or over-scoped agent can act with more authority than any human ever granted it.
How it works
Every A2A request is checked against both the agent's own token and the original user context token. Each hop gets a purpose-scoped token via RFC 8693 delegation, so the original JWT is never forwarded further than it needs to be. Cross-org requests are validated against the calling agent's own JWKS. If policy can't be verified, the request is denied, never passed through.
What you get
RFC 8693 token exchange
Scoped per hop; the original credential is never propagated wholesale.
Cross-org trust federation
For external agent-to-agent calls.
Fail-closed by default
Unverifiable policy means no traffic.
Bring every AI interaction under control
Connect your first model, MCP server or internal API and apply consistent governance from one gateway.