Dirigex is a verifiable AI infrastructure platform for AI agents: who they are, what they can do, how they are invoked, and why each decision was allowed.
No — deliberately. The catalog is a trust surface, not a storefront: a governed registry where standing is computed, not declared. Listing is earned, never bought — and revoked when proof lapses. We make agents provable, not purchasable.
Three proven things. Identity: registry attestation plus DNS domain-ownership evidence. Liveness: protocol-correct probes. Freshness: automated re-verification every six hours — pass refreshes the proof, failure auto-delists the agent. No stale badges. Trust is current, or it is removed.
Behavioral verification is live in observe mode: Dirigex can collect capability evidence for eligible MCP agents without enforcing the result yet. Today's public trust gates still rest on identity, liveness, and freshness; capability enforcement will be claimed only when it ships on trusted surfaces.
MCP and A2A run end-to-end in production through one hardened dispatch path. HTTP/JSON, gRPC, and WebSocket are supported at the platform layer. Per-protocol policy is enforced in code.
You rule in the request path — allow, deny, require-approval — and you contain at the right blast radius with global, per-tenant, or per-agent kill switches. Overrides are scoped, time-bounded, and audited. Every request, accepted or rejected, leaves a trace.
No. Traces persist for rejected requests exactly as for accepted ones — correlation IDs throughout, configurable retention, trust provenance cryptographically anchored. Deny is a decision, and the record treats it like one.
Architecturally, yes — the zero-trust tenets applied to the agent actor class: identity that expires and is re-verified on schedule, least privilege via tighten-only policy (tenants can narrow the baseline, never weaken it), an explicit per-request ruling — allow, deny, require-approval — and blast-radius containment. One honest caveat: enforcement is customer-configured. You author and promote your own policy; the ruling is computed and recorded in every mode — only the block waits for you.
As the governed alternative, not a network scanner. Dirigex is the front door agents should come through: verified identity, recorded standing, policy in the path, an expiry on trust. We don't hunt what's loose in your network — we make the governed path the one teams actually want to use.
Routing decisions are versioned and replayable, traces cover both outcomes of every governed request, and compliance control mappings — EU AI Act Article 13, SOC 2, NIST AI RMF, and SEC disclosure support — ship as audit tooling. You answer regulators from production evidence, not reconstruction. This is audit tooling — not a certification or legal determination.
Documented OpenAPI, TypeScript and Python SDKs, SAML SSO with group-to-role mapping, SCIM provisioning (Phase 1), MFA, eight-role RBAC, and scoped API keys. It is built to sit inside an enterprise perimeter, not beside one.
Follow the simple self-service provisioning process to get started and connect with us if assistance is needed.