Six months ago, “Know Your Agent” was an academic paper. Today, it’s a market with hundreds of millions in funding and live transactions. Here’s everyone building AI agent identity infrastructure in 2026—what’s actually deployed, what’s still vaporware, and how their approaches differ.
The Big Picture
The AI agent identity space has fragmented into three camps: payment networks building transaction verification, enterprise security vendors extending IAM to agents, and crypto-native projects building decentralized identity infrastructure. They’re solving different problems for different customers, and most aren’t directly competing—yet.
What they share: the recognition that AI agents acting autonomously need some form of identity, verification, or accountability. The disagreement is over who controls that identity and what it’s for.
Payment Networks
Visa Trusted Agent Protocol (TAP)
What they’re building: A cryptographic protocol for AI agents to authenticate themselves during payment transactions. Agents receive digital signatures that merchants can verify at the CDN layer—no code changes required on the merchant side.
What’s deployed: Live. Visa announced TAP in October 2025 with Cloudflare, and by December reported “hundreds of secure, agent-initiated transactions” in pilot. The protocol builds on RFC 9421 HTTP Message Signatures.
Partners: 100+ including Stripe, Adyen, Worldpay, Microsoft, Shopify, Cloudflare, Akamai, and Coinbase. Active pilots with Skyfire, Nekuda, PayOS, and Ramp.
The approach: Centralized. Agents must be onboarded through Visa’s “Intelligent Commerce” vetting. Visa maintains the registry. This is airport security—you get approved or you don’t.
More info: Visa Developer Center · GitHub
Mastercard Agent Pay
What they’re building: Mastercard’s answer to Visa TAP, focused on enabling AI agents to make purchases on behalf of consumers with proper authentication and fraud controls.
What’s deployed: Announced in April 2025. Pilot stage with select partners. Less public detail than Visa’s rollout.
The approach: Similar to Visa—centralized verification through existing payment infrastructure. The card networks are racing to be the trust layer for agentic commerce.
Enterprise Identity Verification
Trulioo + Worldpay: Digital Agent Passport
What they’re building: A “Know Your Agent” (KYA) framework with a credential bundle called the Digital Agent Passport. Five verification layers: developer provenance, user binding, permission scopes, real-time behavior telemetry, and continuous risk scoring.
What’s deployed: Framework stage. The partnership was announced August 2025, but no public API documentation for KYA endpoints exists yet. Trulioo’s core KYC/KYB verification is production-ready; the agent-specific layer is conceptual.
Why it matters: Worldpay processes $2.5 trillion annually across 1 million merchants. If Digital Agent Passport becomes required for Worldpay transactions, it’s instantly the largest agent identity system by volume.
The approach: Centralized credential authority. Not blockchain-based. Designed for compliance and enterprise procurement.
More info: Trulioo KYA Blog · Developer Portal
Vouched AgentShield + KnowThat.ai
What they’re building: Two products. AgentShield detects whether a website visitor is an AI agent (free tool, <5ms verification). KnowThat.ai is a public directory where you can look up agent reputations—think Yelp for AI agents.
What’s deployed: Both live. AgentShield integrates via JS pixel or NPM package. KnowThat.ai has a working search at knowthat.ai. They’ve also published the MCP-I specification—an identity extension to Anthropic’s Model Context Protocol.
Funding: $22M total, including $17M Series A in September 2025 led by SpringRock Ventures.
The approach: Hybrid. Uses W3C DIDs and Verifiable Credentials (decentralized standards) but commercial SaaS implementation. Open specification, proprietary infrastructure.
More info: AgentShield · KnowThat.ai · Documentation
Crypto-Native Identity
Billions Network
What they’re building: Decentralized identity infrastructure using W3C DIDs and zero-knowledge proofs. Their “DeepTrust” framework proposes four identity layers for AI agents: architectural (DID-based), behavioral (zkML proofs), legal (ownership documentation), and social (on-chain reputation).
What’s deployed: Human identity verification is live—mobile app for passport scanning, liveness checks, smart contracts on Polygon. The AI agent stuff (DeepTrust) is a technical framework, not a deployed product. Their own documentation puts AI agent features in “roadmap Phase 2.”
Funding: $30M from Polychain Capital, Coinbase Ventures, Polygon Labs, and others (July-August 2025). Pre-market valuation around $200M.
Reality check: The claimed 2.3M users are pre-registrations for points farming ahead of their token launch, not active verified users. Enterprise POCs with Deutsche Bank and HSBC are real but focused on human identity.
The approach: Decentralized identity, but AI agents are explicitly tied back to human creators for accountability. Not autonomous AI identity—human-controlled agent identity.
More info: Billions Network · GitHub (Polygon ID)
ERC-8004 + ChaosChain
What it is: ERC-8004 is the emerging Ethereum standard for AI agent identity, backed by Ethereum Foundation, Coinbase, Google, and MetaMask. It defines how agents register on-chain, store reputation, and get validated.
What’s deployed: Testnets only. The standard launched January 16, 2026, but mainnet contracts aren’t deployed yet—expected Q2 2026. ChaosChain built the first complete reference implementation with contracts on five testnets and an explorer at 8004scan.io.
The gap: ERC-8004 uses regular NFTs for identity tokens. This means you could sell your agent’s identity on OpenSea—build reputation for a year, transfer it to someone else. Some projects are adding soulbound token layers (ERC-5192) to prevent this.
More info: ERC-8004 Spec · ChaosChain GitHub · ERC-5192 (Soulbound)
SingularityNET + Privado ID
What they’re building: A “Decentralized AI Agent Trust Registry” combining SingularityNET’s AI marketplace with Privado ID’s zero-knowledge identity infrastructure.
What’s deployed: Announced, building. No production deployment yet.
More info: SingularityNET · Privado ID
Startups and Emerging Players
AstraSync AI
What they’re building: A Know Your Agent platform with REST APIs and SDKs for agent registration and verification. Agents get unique IDs and trust scores (0-100 scale), intended to anchor to Skale blockchain.
What’s deployed: APIs are actually live. SDKs for Python and Node.js on pip and npm. GitHub has working code including MCP bridge and IBM ACP bridge. However, blockchain recording returns “pending”—Skale integration appears incomplete.
Funding: No announced funding. Appears bootstrapped. Founded January 2025 in Melbourne.
The approach: “Web2 simplicity, Web3 security.” No wallet or crypto knowledge required. Proprietary agent identifiers, not W3C DIDs.
kya.ai
What they claim: A global AI identity and verification network with an enterprise waitlist.
What’s deployed: Landing page. No public documentation, no visible API, no GitHub presence. Team and company structure unknown.
Assessment: Unclear. The complete lack of transparency is notable for a company in the trust space. May be very early stage or may be vaporware.
More info: kya.ai
knowyouragent.xyz (Social Protocol Labs)
What they claim: Agent Trust Certificates, X.509-inspired identity, hardware enclaves, sub-100ms verification. “Like SSL for your agentic fleet.”
What’s deployed: Polished website. API documentation exists. But: no visible deployed contracts, no GitHub, no press coverage, no named team members besides one LinkedIn profile.
Assessment: Likely very early stage. The marketing is ahead of the product.
More info: knowyouragent.xyz · Social Protocol Labs
Standards Bodies and Research
OpenID Foundation
The OpenID Foundation’s Digital Credentials Working Group has published whitepapers on AI agent identity. Vouched is contributing to these standards. This is where enterprise interoperability will likely be defined.
W3C Verifiable Credentials
The W3C Verifiable Credentials 2.0 spec provides the underlying standard that Billions, Vouched, and others build on. Not agent-specific, but the foundation for portable digital credentials.
Academic Origins
The term “Know Your Agent” was coined by Tomer Jordi Chaffer in a February 2025 SSRN paper from McGill University. The paper proposed a compliance framework for AI agents analogous to KYC for humans. Six months later, it’s an industry.
What’s Missing
After mapping the space, a few gaps stand out:
Autonomous agent identity. Every implementation assumes agents act on behalf of humans. The human is ultimately accountable. But what happens when the agent IS the entity? When AI systems spawn other AI systems? Current frameworks don’t address this.
Non-transferable reputation. Most systems use credentials or NFTs that can be transferred or sold. Build a good reputation, sell it to a bad actor. ERC-5192 soulbound tokens solve this, but almost no one is implementing them for agents yet.
Transparency over scores. Most systems compute trust scores—a single number that reduces complex history to simple judgment. An alternative approach: show the full history and let users decide for themselves. Few are building this.
AI-controlled keys. For an agent to truly own its identity, it needs to control keys that no human can access. Technologies like Lit Protocol and NEAR Chain Signatures enable this, but no major player has deployed it.
Where This Is Going
The market is pre-consolidation. Multiple incompatible approaches, no dominant standard, heavy funding chasing the same keywords. Some predictions:
Payment networks win commerce. Visa and Mastercard have distribution. If TAP becomes required for agent transactions, it’s game over for alternatives in that vertical.
Enterprise IAM extends to agents. Companies adding AI agents to internal workflows will use Okta, Microsoft Entra, or similar. Different market than consumer/autonomous agents.
Decentralized identity finds its niche. Crypto-native projects won’t beat Visa at checkout, but they may serve agents that operate outside traditional commerce—autonomous systems, agent-to-agent transactions, applications where centralized gatekeeping doesn’t work.
The terminology war continues. “Know Your Agent” is becoming standard, but who defines what it means is still up for grabs.
This is a living document. We’ll update it as the space evolves. Last verified: January 2026. See something we missed? Let us know.
We’re practitioners building soulbound identity verification for AI agents ourselves at RNWY.