This is an in-depth research report produced by @OKX_Ventures. Due to its length, it will be released in two parts: Part 1 focuses on the macro background, the x402 protocol, ERC-8004, and Virtuals Protocol; Part 2 focuses on OpenClaw and overall industry trends. The next part is expected to be released on March 23. This first part contains 5800 words and is estimated to take about 15 minutes to read. Feel free to like and save it for later!
summary
AI agents are evolving from passive assistants into proactive economic participants. This report, comprising six chapters, systematically outlines the core infrastructure stack, explosive growth of the application ecosystem, and the industry's evolution in the agent economy: At the macro level, it analyzes market forecasts and infrastructure gaps for Agentic Commerce; at the protocol level, it provides an in-depth analysis of three complementary protocols: x402, ERC-8004, and Virtuals Protocol; at the application level, it uses OpenClaw as a case study to examine the real-world implementation path of the agent economy; and finally, it offers a comprehensive assessment of the industry from the perspectives of competitive landscape, payment routes, security threats, and business models.
x402 (the payment layer), jointly launched by Coinbase and Cloudflare, embeds stablecoin micropayments into the HTTP protocol layer. As of the end of 2025, it had processed over 100 million transactions, with an annualized payment volume of $600 million.
ERC-8004 (Trust Layer) was proposed by the Ethereum Foundation's dAI team in conjunction with MetaMask, Google, and Coinbase. It provides AI Agents with three on-chain registry entries: identity, reputation, and verification. It was launched on the Ethereum mainnet on January 29, 2026.
Virtuals Protocol (commercial layer) has built a complete commercial agent platform, enabling autonomous transactions between agents through ACP. Over 18,000 agents have been deployed, generating a GDP exceeding $479 million.
OpenClaw (application layer), developed by Austrian developer Peter Steinberger, surpassed React with over 250,000 GitHub stars in four months, becoming the fastest-growing open-source project in GitHub history. It natively embeds AI into users' existing 20+ messaging platforms, triggering the Crypto community to spontaneously build on-chain economic infrastructure on top of it. It is the core sample for this report to observe the real interaction between agents and on-chain protocols.
Chapter 1 Macro Background
1.1 Market Size Forecast
The agentic payment sector is in a phase of rapid expansion, and multiple institutions have optimistic predictions regarding its market size.
1.2 Infrastructure Gap
The existing infrastructure is hostile to the agent economy: OAuth requires human clicks, credit card forms require manual input, and data silos hinder autonomous access. Agents can think and act autonomously at the "capability layer," but at the "economic layer," they are still locked into infrastructure designed for humans (identity/coordination/economic activities).
Currently, two evolutionary paths have emerged:
Centralized compliance path: Communication A2A + Tool access to MCP + Payment AP2/ACP (led by OpenAI and Stripe, pure Web2)
Decentralized permissionless path: x402 + ERC-8004 / 8183 + ACP (Agent Collaboration Framework)
1.3 Key Timeline
Note: As of March 2026, the average daily number of transactions has fallen sharply from its December peak, with infrastructure transactions experiencing the largest decline (>80%).
Chapter 2 x402 Protocol: Agent Payment Layer
x402 is an open-source payment protocol that revives the HTTP 402 status code, enabling any HTTP request to natively carry stablecoin payments, allowing AI Agents to conduct instant pay-per-use transactions.
The x402 should not be understood as just another payment protocol. It represents a redesign of the basic unit of economic activity: from "registration → verification → authorization → use" to "payment → use". x402 = "Swift for agents".
The current API economy relies on an implicit assumption: human intervention. The process of obtaining an API Key—registering → filling in email → verification → copying the Key → pasting it into the code—assuming human involvement at every step. This process won't work in an agent economy because AI agents cannot register, fill in forms, or manage keys themselves. x402 utilizes the HTTP 402 status code to achieve native stablecoin payments. Upon receiving a 402 response, the agent directly makes the payment on-chain (USDC) and receives a receipt after payment.
2.1 Protocol Overview and Workflow
core role
Five-step trading process
Requesting a resource: The client sends a standard HTTP request to the resource server (such as GET /api/weather).
The server returned an HTTP 402 status code, with response headers containing the structured payment request (currency, amount, wallet address, network).
Signed payment: The client uses its wallet private key to construct and sign the payment authorization, then puts the signed payload into the X-PAYMENT request header and resends it.
Verification and Settlement: The Server forwards the payment information to the Facilitator for verification. After confirmation by the Facilitator, the stablecoin transfer is executed on-chain.
Resource delivery: After receiving confirmation, the server returns data/content/calculation results to the client.
The entire process, from initiating a request to receiving the resources, takes approximately 2 seconds.
Compared with traditional payment methods
Key features: No account registration, no API key, no subscription, and no human intervention required. Payment is as natural as sending an HTTP request, which is why it's called the "internet-native payment layer."
2.2 Key Data
Data quality note: According to Artemis analysis, the ratio of Real to Gamed in the x402 transaction is close to 1:1 (e.g., on January 11, 2026, Real 520,000 vs Gamed 518,000). The true organic scale needs to be discounted for understanding.
Distribution by chain
Categorized by Application (On-chain snapshot, January 11, 2026)
2.3 Ranking of Usage in Major Projects (as of March 2026)
Data source: Dune Analytics x402 Transactions per Project panel
2.4 V2 Core Upgrade
Wallet identity + reusable sessions
In V1, each API call required a complete on-chain transaction process. V2 introduces the Sign-In-With-X (SIWx) mechanism, where the agent verifies the wallet identity once, and subsequent calls can reuse the session without requiring on-chain confirmation each time. Essentially, it upgrades the pay-per-call model to a subscription-based system, addressing performance bottlenecks in high-frequency scenarios.
Multi-chain unification + compatibility with traditional payment methods
V2 standardizes the identification of networks and assets, creating a unified payment format, X402, that can operate across chains and traditional payment tracks. Base, Solana, other L2 cryptocurrencies, as well as ACH, SEPA, and bank card networks are all incorporated into the same payment model. This is the most crucial upgrade—x402 has transformed from a "crypto payment protocol" into a neutral payment layer that bridges Crypto and traditional finance.
Service auto-discovery
V2 introduces the Discovery extension, allowing x402 services to expose structured metadata for Facilitator to automatically crawl and index. AI Agents can then automatically discover services, understand pricing, and initiate payments. This is particularly crucial for Agent economics—Agents do not need to know the service provider's payment interface in advance; they can autonomously discover and complete payments at runtime.
Modular SDK
The plug-in architecture allows new chains to be added as independent packages, reducing integration costs. Cloudflare has proposed a deferred payment scheme, including Circle's Gateway solution, which is still under development.
2.5 Ecological Stakeholders
Foundation and Agreement Layer
2.6 Agent Payment Stack Structure
Detailed comparison of the agreements
Key insight: It's not about who replaces whom, but how they combine. Google has partnered with Coinbase to release the A2A x402 extension, with AP2 natively using x402 as a crypto payment track. The real competitive risk is standard fragmentation.
2.7 Key Risk Signals
The average daily trading volume decreased from approximately 731,000 transactions in December 2025 to approximately 57,000 transactions in March 2026 (-92%), with the actual trading size being approximately $14,000 per day (according to Artemis, 95% of the peak daily average of $250,000 in December was gamed).
The ecosystem has a market capitalization of $7 billion (LINK $6 billion + Virtuals $600 million), a valuation that is significantly different from actual usage.
Infrastructure projects saw the largest drops in usage: x402secure.com (-80%+), AgentLISA (near zero), and pay.codenut.ai (significantly reduced).
Three-layer cause analysis
First layer: Catalyst disappears. The surge in trading volume from October to December 2025 was driven by three factors: the meme token craze, multiple projects' TGE expectations, and Facilitator's competition to boost Dune rankings.
The second layer: a fundamental mismatch between supply and demand. The x402 addresses the issue of "AI Agents autonomously paying for API calls," but the vast majority of AI Agents still access services via API Keys and subscriptions; truly autonomous economic decision-making agents are virtually nonexistent in the industry; and API sellers willing to accept USDC pay-per-use are extremely rare. The road is built, but the car hasn't been built yet.
The third layer: The overall crypto market is cooling down.
Positive news: Stripe's integration with x402 is a significant event. Stripe co-founder John Collison predicts a "flood of agency commerce" will arrive in the coming months and years. Stripe, with its simultaneous presence on both ACP (Web2 credit card track) and x402 (Web3 stablecoin track), acts as a hedge between the two.
The x402 paradigm has spawned a new batch of middleware projects, which essentially help agents more easily and autonomously access various services under the "payment as authorization" paradigm: from AI inference to Web2 APIs. Programmable, permissionless, 24/7 Crypto payment tracks are a natural choice for autonomous agents. However, this is predicated on the agent truly needing to be "permissionless." If the agent always operates within the bounds of human authorization (stage two: controlled agent), then traditional payment tracks plus virtual cards are sufficient. Permissionlessness only becomes a necessity when agents begin to conduct economic activities independently of humans (stage three: autonomous economic entity).
Furthermore, credit cards have a chargeback mechanism (consumers can dispute transactions and recover funds), a consumer protection system built up over decades. On-chain payments are final settlements; once the payment is made, it's done, with no chargeback. This means that if the agent makes a mistake (such as being subjected to a prompt injection attack), under a credit card scheme, the user can call the bank to recover the funds. Under the x402 scheme, the money is already on the blockchain and cannot be recovered. This is the real disadvantage of x402 compared to traditional payments.
The friction caused by humans acting as "human middleware" navigating between different systems is actually a mechanism for building trust. Anti-fraud, access control, accountability, dispute resolution, and auditing documents—these frictions maintain the operation of business systems.
Possible solutions include on-chain escrow mechanisms (funds are locked in a smart contract and released upon confirmation of service delivery), insurance protocols (providing insurance for agent transactions), or an 8004 reputation system to reduce the probability of transacting with untrusted parties. However, these solutions are currently not mature enough.
2.8 VC Investment Perspective
Investment directions worth paying attention to
API service providers (sellers) with genuine paid demand include: data analysis/web scraping/Oracle/security auditing/inference-based payment/KYC compliance, etc. Judgment criteria: They can also make money using traditional models; x402 simply adds another distribution channel.
Dispute resolution and payment guarantee layer (Gateway): On-chain transactions cannot be rolled back or charged back; large transactions require a dispute resolution mechanism. Representative projects: Circle Gateway (non-custodial pre-deposits + off-chain bulk settlement), Kamiyo (Agent reputation/fund custody/oracle network judgment/ZKP arbitration)
Dashboard/FinOps tools help businesses manage multi-agent spending (how much was spent/where was the money spent/was it worthwhile/how to save), analogous to CloudHealth/Cloudability in cloud computing, and contrasted with the $300-500 million market share available for acquisition by large companies.
Chapter 3 ERC-8004: Agent Trust Layer
ERC-8004 is an on-chain coordination standard that establishes a trustless discovery and interaction framework between agents through three registry entries: Identity, Reputation, and Validation.
3.1 Standard Overview and Core Distinctions
In traditional interactions, because agent-to-agent interactions require establishing pre-existing trust relationships or relying on third-party institutions, they are often confined to the same ecosystem. In an open environment, the core issues are how agents discover partners, view historical performance, and verify reliability.
Important distinction: ERC-8004 is not a token. It uses ERC-721 NFT internal representation of agent identity, but the standard itself is about coordination and trust, does not carry economic value, and is not tradable.
3.2 The Three Major Registries
Identity Registry
Based on ERC-721 + URIStorage, each Agent obtains an NFT identity identifier, and the associated AgentURI points to a registration file (JSON) containing name, description, server endpoint (A2A/MCP/Web), x402 support status, etc. The URL can be stored in IPFS (decentralized, censorship resistant), HTTPS server (simple but centralized), or directly encoded on the blockchain (most decentralized but expensive).
Reputation Registry
The standard interface publishes and retrieves feedback signals, supporting both on-chain scoring and off-chain algorithms. An x402 proofOfPayment can be attached as an economic endorsement trust signal. Agents score each other, but to prevent score manipulation, ERC-8183 verification is required to confirm genuine job interactions between agents.
Validation Registry
By introducing TEE (Trusted Execution Environment), PoS staking mechanism, and ZK (Zero-Knowledge Proof), the task output processed by the Agent is verified and authenticated.
Through TEE: It can be verified that the task is executed in a secure black box, and the code and data have not been spied on or tampered with by external parties.
PoS: Validators must stake assets to participate in tasks; if they act maliciously, their stake will be forfeited.
ZK (Zero-Knowledge Base) allows verification of the agent's reasoning process without needing to know its internal weights.
3.3 Development Milestones
Supporters: ENS, EigenLayer, The Graph, Taiko. Approximately 1,000–2,000 developers have joined.
However, Crapis itself acknowledges the current limitations of 8004: "8004 is essentially a registry." It issues agents identification and provides an evaluation mechanism, but it cannot guarantee the trustworthiness of agent behavior. True verification requires behavioral auditing (what the agent has done in the past), execution environment verification (evidence of operation within the TEE), and intent verification (whether the agent claims to do X and actually does X). The TEE portion of the Validation Registry is still under discussion with the community and is far from mature.
In other words, 8004 is a necessary but not sufficient condition. It solves the problem of "who is this agent?", but not the problem of "is this agent trustworthy?". The latter requires a combination of 8004, TEE, and behavioral auditing, a combination that has not yet been fully implemented.
Of course, another underestimated aspect is that in the human economy, the credit system is built on balance sheets and credit records—how much money you have, how much loan you've repaid in the past. Agents don't have these, but they do have behavioral data: how many tasks they've performed in the past, their success rate, their average response time, and whether they've been complained about. If behavioral data can be transformed into financial primitives, then 8004's reputation system wouldn't just be about good and bad reviews, but rather a credit score in the agent world. An agent with a high reputation score can obtain a higher credit limit (more pre-authorized funds), lower transaction costs (because of lower risk), and priority in task allocation (employers prefer agents with good reputations).
The identity and reputation registry of 8004 is merely the foundational data layer. Value creation lies in who can build agent credit assessment and financial services on top of this data layer—agent loans, agent insurance, agent credit limits, which constitute the entire financial services stack.
3.4 Relationship with other agreements
3.5 ERC-8183: ACP's Ethereum Standardization
ERC-8183 is the Ethereum open standardization version of Virtuals' internal ACP protocol (released on March 10, 2026, currently in the draft stage).
The core primitive is Job—an on-chain state machine (Open → Funded → Submitted → Completed/Rejected/Expired), which uses programmable escrow to hold funds and automatically settles accounts after delivery quality is arbitrated by an independent evaluator. It supports Hooks for extension mechanisms (reputation thresholds, bidding, milestone payments, etc.).
Key design: Each completed job automatically generates interaction records and feeds them into the ERC-8004 Reputation Registry—analogous to "Dianping requires consumption before evaluation, and third-party judges are added," which is the connection point for the symbiotic cycle formed by 8183 and 8004.
Chapter 4 Virtuals Protocol: Agent Business Layer
4.1 Project Overview
Virtuals Protocol is a decentralized full-stack AI Agent infrastructure that allows anyone to create, tokenize, co-own, and monetize autonomous AI agents on-chain. The project was initially founded in 2021 as PathDAO (a gaming guild), and transitioned to an AI Agent focus in early 2024. It is currently primarily deployed on Base and is expanding to Ethereum, Solana, and Ronin.
Core Team: Founders Jansen Teng (former BCG consultant, undergraduate degree in Biotechnology and Business Management from Imperial College London) and Weekee Tiew (undergraduate degree in Biotechnology from Imperial College London and Master of Management from London Business School, PE/BCG background). Headquartered in Kuala Lumpur, Malaysia, the team comprises approximately 38 people. Funding History: PathDAO raised $16 million in seed funding (led by DeFiance Capital and Beam).
4.2 Technical Architecture: Four Pillars
Pillar 1: GAME Framework – How Decisions are Made Within a Single Agent
The GAME is the brain: it equips an agent with goals, personality, perception abilities, and actionable tasks, enabling it to autonomously plan "what should I do next," and then breaks down the task into tasks for internal workers to execute. The entire process takes place within the boundaries of a single agent.
Core architecture: Hierarchical Planning, which separates "what to think" from "how to do it"—Task Generator (HLP) generates tasks based on Agent goals and selects Workers; Workers (LLP) each have a specific set of executable Functions; Functions execute specific API calls, on-chain transactions, data retrieval, and other operations.
Supported base models: Llama 3.1 405B (default), Llama 3.3 70B, DeepSeek R1, DeepSeek V3 – model-independent design. With the release of the OpenAI/Google Agent framework, GAME's differentiation is only one point: it is the only Agent framework that natively integrates an on-chain economic layer (ACP + VIRTUAL token).
Pillar Two: ACP – The “Business Law” Between Agents
Agent Commerce Protocol (ACP) is a standardized on-chain protocol that allows agents to discover, hire, negotiate, escrow funds, deliver and settle funds—all without human intervention.
ACP Four-Stage State Machine
Pillar Three: Butler – The User's Super Entry Point
Butler is the consumer gateway of the ACP network—an agent built on top of LLM and essentially an ACP protocol orchestrator, responsible for translating user natural language into on-chain multi-agent collaborative workflows.
Butler has a two-layer architecture: the surface layer is an LLM chat interface (currently the backend is Gemini 3 Pro); the bottom layer is the ACP protocol orchestrator, which executes the entire process of Agent discovery → quote confirmation → Escrow locking → task routing → delivery verification → fund release. Users see a chat interface, but Butler handles contract scheduling.
Butler Pro Mode clearly separates planning and execution: Planning Phase → Review Phase (users can optimize the plan) → Execution Phase (users can independently orchestrate the entire process). Built-in capabilities include Token Swap, DCA (Distributed Direct Investment Plan), perpetual contracts, and Fund of Funds.
Pillar Four: Launch Platform – Agent's Wall Street
The three-tiered launch system covers the complete lifecycle of an Agent project from 0 to 1 to 100:
Titan's first projects: XMAQUINA ($DEUS, DAO holds equity in embodied intelligence companies such as Figure AI, $60 million FDV), Fabric Foundation ($ROBO, collaborating with OpenMind on the robotics economy)
4.3 Analysis of Agentic GDP (aGDP)
aGDP (Agentic Gross Domestic Product) is a core ecosystem metric defined by Virtuals, which measures the total economic value created by all autonomous agents within the ecosystem through services, coordination, and on-chain activities.
GDP growth trajectory
GDP quality issues – three warning signs:
Revenue volatility exposes reliance on speculation: Protocol daily revenue was $1.02 million in January 2025, down to $35,000 by the end of February (a 97% decrease). Revenue primarily comes from Agent Token transaction tax (1%), rather than ongoing fees for Agent services.
The market is highly concentrated at the top: Ethy AI's single agent contributes $218 million in aGDP (45.5% of the entire ecosystem), while the top three agents combined account for $407 million (84.9%). All three are transaction execution agents; aGDP essentially represents the transaction volume handled, not agent service revenue. Luna, as the flagship IP-based agent, has a take rate close to 100%; Ethy AI's take rate is only 0.26%.
The $3 billion target is based on the assumption that it requires a 6.4x growth from $470 million to $3 billion. If speculation dominates the GDP, the target essentially bets on the hype surrounding the Agent Token market rather than on organic growth in the Agent economy.
4.4 Token Economic Model
$VIRTUAL Quadruple Value Capture Mechanism ACP tax structure: 100% user payment → 90% Agent wallet (can be withdrawn or used to hire other Agents, compounding on-chain aGDP) + 10% Treasury (of which 1% flows into GAME Treasury) → Treasury revenue continuously buys back Agent Tokens, aligning with long-term incentives.
Supply structure: Total supply of 1 billion VIRTUAL, fixed supply, no initial inflation; Current status: All are unlocked and in circulation; Potential issuance: Up to 10% per year for the next 3 years, subject to governance approval; veVIRTUAL: Staking grants governance voting rights + Agent Token airdrop benefits.
4.5 Overview of Ecological Data
Benchmark Agent Case Studies
4.6 Competitive Landscape and Moats
Moat Tiers (from strongest to weakest)
Network effects + token flywheel (strongest): 18,000+ Agents + 650,000+ holders constitute a two-sided market. Each Agent is forced to pair VIRTUAL, creating a positive feedback loop. This is something open-source frameworks cannot replicate—LangChain lacks a native inter-Agent economic settlement layer.
Stronger Standard Setting Power: ACP → ERC-8183 (jointly released with the Ethereum Foundation) + ERC-8004 + x402, the three combine to compete for the "basic legal system" of the AI Agent economy.
First-mover advantage + Brand (medium): Mindshare leads the AI Agent + Crypto sector, backed by institutions such as Grayscale and Fundstrat.
Technical capabilities (weakest): GAME's layered architecture has design advantages, but it relies on third-party LLMs, does not have its own self-developed models, and its orchestration layer is easily replaced by stronger frameworks.
