The problem. Enterprises are deploying AI agents faster than they can govern them. Gartner predicts over 40% of agentic AI projects will be cancelled by end-2027 due to escalating costs, unclear value, or inadequate risk controls. The reason: agents are "ghost identities", over-privileged, reused, and untraceable, and native cloud IAM stops at the platform boundary, leaving no way to answer an auditor's simplest question: who approved this agent's action, and on whose authority?

What Agent Gateway does. Agent Gateway is the flagship product in Affinidi's Trust Fabric suite, one control point that every agent request routes through, checked at runtime along four dimensions: Who (verified identity and credentials), What (tool, dataset, or service being accessed), When (session state, rate limits, business hours), and Why (declared, logged business purpose).

In practice, that means:

Identity at birth. Every agent gets a W3C-standard, cryptographically verifiable DID the moment it's created, with no code rewrite.
Caller context binding. Every downstream action is traceable back to the human or system that authorised it.

Runtime policy enforcement. Fine-grained, OPA-based rules (drag-and-drop or JSON) gate every request, not just at deployment time.

Credential injection. Secrets and credentials are enriched into requests centrally; the agent itself never holds the keys.

Protocol-aware proxy. Drop-in support for A2A, MCP, and REST-to-MCP hosting, so legacy systems and any agent framework (LangGraph, CrewAI, AutoGen) plug in without replatforming.

Why it matters for regulated markets. Singapore's IMDA Model Governance Framework for Agentic AI (v1.5) and MAS's Safeguards for Agentic Finance at Runtime both call for exactly this: a unique agent identity tied to a supervising human, human-approval checkpoints, and a complete audit log of every agent-to-agent step. Agent Gateway operationalises that requirement directly, rather than asking enterprises to bolt governance on after the fact.

Who it's for: Platform and infrastructure leaders, CISOs, and compliance teams at any organisation running (or about to run) AI agents in production, particularly in financial services, healthcare, and cross-border data exchange, where accountability and auditability aren't optional.

Affinidi is Temasek-founded, ISO/IEC 27001 certified, and built entirely on open standards (W3C, DIF), so the governance layer you adopt is portable, not another walled garden.