GoDaddy, Infoblox, Google and Cloudflare have demonstrated an open reference architecture designed to let artificial intelligence agents identify and authenticate one another while completing transactions across separate online platforms.
The companies unveiled the Chain of Trust at GovWare 2026 in Singapore. The project was developed with Singapore’s Infocomm Media Development Authority (IMDA) and applies principles from the regulator’s Model AI Governance Framework for Agentic AI.
The demonstration simulated the purchase of a flight from Madrid to Singapore by agents that had not interacted before and were running on different infrastructure. The buyer agent operated on Amazon Web Services, the seller ran on Cloudflare Workers, Google provided mandate and verification components, and Stripe handled payment.
Open standards connect the transaction
The architecture links several existing and emerging standards rather than relying on a single vendor’s closed identity system. Agent Resources Discovery and DNS-based lookup support discovery; the Agent Name Service and SPIFFE provide identity evidence for agents and running instances; and OAuth is used to define authorization.
Agent2Agent and the Model Context Protocol support communication between agents and tools. The Agent Payments Protocol and Universal Commerce Protocol carry a person’s signed purchasing authorization and the agent’s identity into checkout. Supply Chain Integrity, Transparency and Trust provides an append-only record, while a Trust Index converts attestations into an evidence-based score.
The project’s technical paper presents those layers as independently verifiable controls. That separation is important because passing one check should not allow an agent to bypass the others. It also gives businesses and governments a way to set their own trust policies rather than accepting a platform’s internal decision.
The demo tested identity and authorization
In the demonstration, the buyer agent searched for a flight, discovered a seller and checked the seller’s identity before proceeding. The architecture was also tested against a look-alike seller and forged credentials. Those attempts were rejected, according to the joint announcement.
The user approved the final purchase with a passkey and a signed payment mandate. That mandate specified what the agent was authorized to buy and allowed the merchant and payment provider to check the transaction against the user’s instruction.
The approach addresses a growing interoperability problem in agentic commerce. If each platform verifies agents only within its own system, a merchant may have little evidence about an agent arriving from another cloud or protocol. A portable identity and audit trail can make the basis for trust visible across organizational boundaries.
The design also creates several policy choices for adopters. Organizations must decide which identity issuers, attestations and trust thresholds they will accept, how long credentials remain valid and what evidence is retained. Open protocols make those decisions portable, but they do not remove the need for risk management.
A reference architecture, not a production guarantee
The Chain of Trust remains a reference architecture demonstrated in a controlled scenario. The announcement does not establish that the system has been deployed at commercial scale, independently certified or adopted as an industry standard. Production use would still require implementation work, governance decisions, security testing and agreement among participants on how trust signals are evaluated.
IMDA’s involvement gives the work a direct connection to Singapore’s agentic-AI governance program. The authority’s framework emphasizes meaningful human accountability, technical controls and transparency as AI systems take on more autonomous tasks. The transaction demo turns some of those principles into a concrete workflow spanning discovery, identity, authorization, communication, payment and audit.
The broader significance is the attempt to make agent identity portable across platforms. Whether the architecture gains traction will depend on support from merchants, payment providers, cloud platforms and standards bodies, as well as evidence that the controls remain reliable under production conditions.
Featured image: Immo Wegmann on Unsplash
Singapore launches New Model AI Governance Framework for Agentic AI