Three distinct blockchain-based identity and trust frameworks for autonomous AI agents launched or expanded in a single seven-day window ending April 23, 2026. Zetrix AI and China's CAICT unveiled the Avatar trust protocol on April 14. 0G Foundation partnered with Alibaba Cloud on April 21 to bri...
"At a high level, inference runs on Qwen, and verification runs on 0G — forming a more complete foundation of compute and trust for autonomous AI systems." — Michael Heinrich, CEO and Co-Founder, 0G Labs
Three distinct blockchain-based identity and trust frameworks for autonomous AI agents launched or expanded in a single seven-day window ending April 23, 2026. Zetrix AI and China's CAICT unveiled the Avatar trust protocol on April 14. 0G Foundation partnered with Alibaba Cloud on April 21 to bring Qwen large language models on-chain with verifiable compute. Coinbase's Agentic.Market went live on April 20-21 atop the x402 payment protocol, which now reports 167 million settled transactions and 480,000 transacting agents. In parallel, HashKey Group published a whitepaper on April 21 proposing a dual-token architecture — AI tokens for compute, blockchain tokens for value transfer — as the financial plumbing for what it calls the "agent economy era."
The convergence is not coincidental. With 45,000+ agents registered under the ERC-8004 identity standard on Ethereum mainnet since January 2026, and the decentralized identity market projected at $7.4 billion this year, the race to define how autonomous software authenticates, transacts, and builds reputation on-chain has entered a standards war phase. The outcome will determine which trust layer captures the coordination fees of a market that Gartner estimates could influence $30 trillion in annual machine-initiated purchases by 2030.
AI agents that hold wallets, execute transactions, and interact with smart contracts present a verification problem that did not exist when blockchains served only human users. A human signs a transaction with a private key and implicitly stakes their reputation. An autonomous agent operating at machine speed — executing more transactions per hour than a human power-user does per month, according to Cobo's technical documentation — needs a different trust model. The agent must prove what it is authorized to do, who it represents, and whether its credentials are valid, all without stopping to ask a human.
The a16z crypto research team articulated the structural case in February 2026: the internet lacks a native mechanism to separate humans from machines while preserving privacy and usability, and blockchain-based identity provides portable, universal "passports" for agents that are difficult to spoof. Traditional payment systems are a bottleneck for machine-scale transactions. Existing financial infrastructure assumes physical accounts, manual audits, and staged clearing — none of which scale when agents negotiate and settle autonomously.
This week's launches represent three geographically and philosophically distinct answers to that problem.
Launch date: April 14, 2026, at the World Internet Conference Asia Pacific, Hong Kong.
Zetrix AI Berhad (formerly MY E.G. Services Berhad) and the China Academy of Information and Communications Technology (CAICT) unveiled the Avatar protocol, a blockchain-based trust layer for what they term the "agentic AI ecosystem." Avatar runs on a hybrid architecture integrating the Astron blockchain (CAICT's initiative) and the Zetrix layer-1 chain.
The protocol enables individuals and enterprises to create AI "digital twins" trained on specific personalities, knowledge bases, and communication styles. These agents can then interact with other agents, users, and systems while carrying verifiable credentials — identity documents, professional qualifications, and access to digital assets — anchored on-chain.
Key technical features, according to the press release:
"Blockchain-empowered agentic AI will form the trust foundation for next-generation autonomous digital interactions," said You Xiao Yu, Vice President of Astron CAICT.
TS Wong, Group Managing Director of Zetrix AI, framed the commercial rationale: "Agents representing users herald a new wave requiring knowing who you transact with and accessing verifiable credentials."
Avatar's geographic positioning matters. With CAICT — a scientific research institute under China's Ministry of Industry and Information Technology — as a co-developer, the protocol is designed to operate within China's regulatory framework. This positions it as the credentialed identity layer for AI agents in the Chinese market, where the government has mandated AI registration and labeling requirements.
Announcement date: April 21, 2026.
The 0G Foundation partnered with Alibaba Cloud to bring the Qwen family of large language models on-chain for the first time. Qwen3.6, Alibaba's latest release, is described as one of the most widely downloaded model families globally, supporting multimodal and enterprise use cases.
The integration enables a specific capability: AI agents can query Qwen to generate responses, process information, and execute multi-step tasks in real time, with all inference verifiable through 0G's data layer. This means developers can build workflows where the AI reasoning is auditable — a 100% auditable agentic pipeline, per the announcement.
The architecture separates concerns: inference runs on Qwen's compute, verification runs on 0G's blockchain layer. According to Michael Heinrich, CEO of 0G Labs, this forms "a more complete foundation of compute and trust for autonomous AI systems."
The practical shift is from centralized API access — requiring account setup, fiat billing, and manual configuration — to tokenized, programmatic access. Agents pay per inference call using tokens rather than credit cards. This removes the API key dependency that has historically bottlenecked autonomous agent operations.
0G's positioning in this market is infrastructure-layer: the Foundation describes itself as providing "Blockchain for AI Agents" and has invested in grants, ecosystem development, and community governance to attract developers. The Alibaba Cloud partnership gives it access to enterprise-grade LLMs without building its own models.
Launch date: April 20-21, 2026.
While Avatar starts from identity and 0G starts from compute verification, Coinbase's approach starts from payments. The x402 protocol, now governed by the Linux Foundation with backing from Google, Microsoft, AWS, Visa, Mastercard, Stripe, and Shopify, embeds stablecoin micropayments into HTTP requests. AI agents pay for services per call, with no API keys or subscriptions required.
Agentic.Market, the marketplace layer launched this week, organizes x402-enabled services across seven categories: inference, data, media, search, social, infrastructure, and trading. Providers include Bloomberg, AWS, CoinGecko, Anthropic (ranked #1 by agent spending), Google Flights, and OpenAI.
Coinbase product lead Nick Prince described the goal: "Give humans and their agents access to thousands of services, with zero API keys required."
Cumulative x402 metrics as of April 21, 2026:
These numbers require the same scrutiny applied in prior analysis. An average daily volume of approximately $28,000 against billions in ecosystem token valuations indicates the market is still in an early deployment phase. The x402 protocol itself captures zero fees — value accrues to services transacted over it, not to the rail.
Proposed in August 2025 by contributors from MetaMask, the Ethereum Foundation, Google, and Coinbase, ERC-8004 went live on Ethereum mainnet on January 29, 2026, and has since deployed across 20+ networks. The standard provides three on-chain registries:
As of April 2026, more than 45,000 agents are registered. A v2 specification is in development with enhanced MCP (Model Context Protocol) support and x402 integration. The standard extends Google's Agent-to-Agent (A2A) protocol with a trust layer for cross-organizational agent interaction.
ERC-8004 occupies a different layer than Avatar or 0G. It does not provide compute or credential verification from a government authority. It provides minimum viable trust infrastructure — identity, reputation, validation — as a public good. Application-specific logic stays off-chain.
On April 21, 2026, HashKey Group Chairman and CEO Dr. Xiao Feng released the third installment in the Web3 Economy whitepaper series: "On-Chain Finance and Tokenization in the Era of the Agent Economy." The paper proposes a structural framework for how value should flow in agent-mediated economies.
The core contribution is a Dual-Token Architecture:
The paper argues that traditional financial systems assume physical accounts, manual audits, and staged clearing — infrastructure designed for human-speed commerce. When agents operate at scale, these assumptions break. The whitepaper identifies three tokenization advantages for agent economies: granularity (splitting assets and permissions into smaller units), automation (smart contracts encoding trading rules), and unified ledger (reducing intermediary steps).
HashKey also proposes expanded use of Soulbound Tokens (SBTs) for agent identity, permission management, and reputation — conceptually overlapping with ERC-8004's reputation registry but framed within a broader economic model.
| Feature | Avatar (Zetrix/CAICT) | 0G + Alibaba | x402 / Agentic.Market | ERC-8004 | |---|---|---|---|---| | Primary function | Identity + credentials | Compute verification | Payments | Identity registry | | Blockchain | Astron + Zetrix | 0G chain | Base (Ethereum L2) | Ethereum + 20 chains | | Geographic focus | China / APAC | Global | Global (US-led) | Global | | Government backing | CAICT (China MIIT) | None | Linux Foundation | Ethereum Foundation | | Agent count | Not disclosed | Not disclosed | 480,000+ | 45,000+ | | Revenue model | Task store fees | Token-gated compute | Zero protocol fees | Public good | | Credential type | Government-certified | Verifiable inference | Payment history | On-chain reputation |
The table reveals a market segmenting along jurisdictional and functional lines. Avatar targets regulated credential verification in China. 0G targets verifiable AI compute globally. x402 targets frictionless payments. ERC-8004 targets composable identity. No single standard covers the full stack.
The economic question is where value accrues in a stack where agents need identity, compute, payment rails, and reputation simultaneously.
The x402 protocol has explicitly chosen not to extract rent — zero fees, open specification. Value flows to service providers: Anthropic, Bloomberg, and AWS rank among top earners on Agentic.Market. This mirrors traditional internet economics where protocols (TCP/IP, HTTP) are free and applications capture value.
Avatar's task store model introduces platform fees, creating a more traditional marketplace dynamic. The CAICT certification requirement functions as a barrier to entry, potentially concentrating value among certified credential providers.
0G's model gates access through token holdings, meaning the $0G token captures value proportional to demand for verifiable AI inference. This is a direct bet that auditability commands a premium over standard API access.
ERC-8004, as a public good with no fee extraction, creates value for the applications built on top of it rather than for the standard itself. The 45,000 registered agents represent potential demand for reputation scoring, validation, and identity resolution services — all of which can be monetized at the application layer.
The total addressable market is significant but uncertain. The decentralized identity market is projected at $7.4 billion in 2026, according to industry research. The EU's eIDAS 2.0 regulation requires member states to deploy digital identity wallets by end of 2026, potentially creating regulatory demand for standards-aligned agent identity systems. A PwC survey of over 300 companies found 79% are already adopting AI agents in some form.
The simultaneous emergence of multiple AI agent identity and trust protocols in April 2026 marks the transition from theoretical agent economies to a standards competition with real infrastructure deployment. The parallel is not to previous "blockchain killer" narratives but to the browser wars or mobile OS competition — races where the winning standard captured coordination value for decades.
The data suggests the market is not yet large enough to support multiple competing trust layers. x402's $28,000 daily volume, split across 480,000 registered agents, translates to less than six cents per agent per day. ERC-8004's 45,000 registered agents represent a fraction of the autonomous software already deployed in enterprise settings. The infrastructure is being built on the expectation that agent transaction volumes will follow deployment, not the reverse.
What differentiates this standards race from prior blockchain competitions is the involvement of non-crypto institutions. CAICT is a Chinese government research body. The x402 Foundation includes Visa, Mastercard, and Google. Alibaba Cloud is one of the world's largest cloud providers. The entry of these institutions signals that AI agent trust infrastructure is not perceived as a crypto-native problem but as a general internet infrastructure requirement — one where blockchain happens to provide the most viable solution for machine-readable, permissionless identity and micropayment settlement.
The coming quarters will determine whether the market consolidates around one or two standards or fragments along jurisdictional lines, replicating the regional internet fragmentation patterns already visible in data governance and AI regulation.