Early Access
A coordinated system for agentic trust
A control layer for AI agents: cryptographic identity, policy enforcement, and live observability across calls, hops, and organisational boundaries.
Components
Three components, one shared trust model. Start with Agent Gateway today and extend to Stream and Pay without reworking identity or policy.
Available now
Agent Gateway →
Where agent interactions are authenticated and policy-enforced.
- Identity you can prove, on every request.
- Policy is enforced inline, at the moment of action.
- Works across protocols: MCP, A2A, AP2, x402.
Coming soon
Agent Stream
Where AI execution flows are governed in real time.
- Unified execution layer across models, tools, and databases.
- Continuous evaluation and routing with automatic failover.
- Built-in safeguards at every step.
Coming soon
Agent Pay
Where AI transactions are settled in real time.
- Enforce payment conditions before access proceeds.
- Confirm payment proofs and handle settlement under policy.
- Support payment policies using x402 (HTTP-native payments) and MPP (multi-party payments).
Component architecture
All agent calls enter through Agent Gateway. Agent Stream and Agent Pay are called through the gateway. The Agent Caller has no direct path to Agent Pay.
Agent Gateway
Authenticates the caller, attaches a verifiable identity, and evaluates OPA policies before any backend is touched.
Agent Stream
Routes calls across providers through one API, applies guardrails, and handles automatic failover. All requests remain bound to the same identity context.
Agent Pay
Enables agents to authorise and settle services in real time under the same policy-bound trust controls.
Get started
Configure a surface on Agent Gateway and make your first governed request.
Learn how identity and policy work before you start configuring.
Configure authentication, define policies, set up observability.