x402 Since Launch: 150 Million Transactions, Settlement Amounts, and a Review of Real-World Use Cases

Author: Yuki (Liu Yuqing) Stablehunter/Money in Motion

On May 6, 2025, the Coinbase Developer Platform officially launched x402, a protocol that revives the long-forgotten 402 Payment Required from HTTP, allowing software to receive payment requests directly when requesting resources, complete stablecoin payments, and then obtain API, data, content, or tool invocation results.

More than a year later, we finally have some data to discuss. According to public information, under different statistical criteria, x402 has already shown three levels of adoption signals:

These numbers can no longer be seen as just a protocol demo, but that does not mean that Agentic Payment has exploded; rather, we need to examine what these transactions truly represent.

AI Agents are really starting to take over tasks for humans. When an agent calls a paid API, how does it make the payment? If an agent temporarily purchases a piece of data, how does it settle that? If an agent only needs to use a certain MCP tool once, does it still need to register an account, bind a credit card, and activate a subscription? x402 attempts to answer these types of questions.

What is the 150 Million Transaction Data of the x402 Protocol?

Many people see the data for x402 and their first reaction is: 150 million transactions, 50 million USD in settlement amounts, does this indicate that Agentic Payment has entered large-scale adoption?

This indicates that x402 is already operational, but it does not yet prove that real commercial demand has fully established itself. Chainalysis mentioned in its June 2026 analysis that the cumulative transactions of x402 on Base have exceeded 100 million, which is an important milestone because Base is one of the earliest core networks for x402.

However, a significant portion of the early growth came from activities like PING, where users obtained some form of participation credentials or social benefits through extremely low-cost on-chain actions. This is not the same as enterprise agents continuously purchasing services for real workflows. Therefore, 100 million transactions prove that the protocol has sufficiently low friction to support a large number of high-frequency, small-scale, automated interactions. It has not yet fully proven that these interactions represent sustainable commercial demand.

Looking at the cross-chain tracker, the x402 tracker for the Agent Economy shows that as of around July 10, 2026 UTC, there have been approximately 155.6 million cumulative x402 transactions and approximately 40.97 million USD in settled stablecoin volume, covering multiple chains and facilitators.

Another metric worth looking at separately is x402scan. When I checked the x402scan homepage data on July 11, 2026, there were approximately 16.24 million transactions, about 837,500 USDC in total amount, about 67,000 unique buyers, and about 39,600 unique sellers within the current index range. This number is significantly lower than the total cross-chain volume of the Agent Economy, so I would not consider it as the "total transaction volume of the entire network x402".

It is more suitable for answering another question: among the visible sellers/origins in x402scan, who is really generating transactions continuously? The number of transactions represents the frequency of calls, while the settlement amount represents the scale of funds. The number of active agents represents participating entities, while the number of facilitators and chains represents the diffusion of infrastructure.

x402 has gradually transformed from a protocol initiated by Coinbase into an early payment layer across networks, service providers, and developers. But every number requires us to ask: how many of these transactions come from real agent workflows? How many come from developer testing? How many come from incentive activities, stress tests, bot behaviors, or one-time experiments?

If we do not ask this question, the data for x402 can easily turn into another illusion familiar to the Crypto world: the numbers look good, but we do not know if there is real demand behind them.

The Real Value of x402 Lies Not in Amounts, but in Transaction Forms

I believe the most important number for x402 right now is not the settlement amount, as the tens of millions of USD in stablecoin volume is not significant in the entire payment market. Compared to networks like Visa, Mastercard, Stripe, and PayPal, it is still at a very early experimental scale.

However, the value of x402 does not lie in how much has been settled now, but in that it has allowed a new form of transaction to begin to emerge: software can see prices, complete payments, and gain access to resources all in one HTTP request. For human users, this may just mean one less checkout page.

But for agents, it is important because agents are not suited to follow human-designed business processes every time: registering accounts, verifying emails, binding credit cards, selecting packages, copying API keys, and waiting for approval.

The behavior of agents is more like discovering external capabilities continuously during the task completion process: if it needs a search, it buys a search; if it needs a data point, it buys a data point; if it needs to call a model, it pays for that model call; if it needs a security check, it purchases an audit; if it needs access to a paid MCP tool, it completes authorization and settlement within the task context. This is not the logic of traditional SaaS subscriptions. This is the logic of purchasing by task, by call, and by capability. Therefore, for x402, the number of transactions is more worthy of attention than the settlement amount.

If a payment protocol is truly to serve agents, it should not only look at GMV early on, but should focus on whether small, high-frequency, automated calls can be established.

An agent paying 0.001 USD for an API call is a small amount, but it represents a significant commercial action: for the first time, machines can purchase external capabilities to complete tasks without human checkout.

What Are the Real Use Cases?

Currently, the use cases for x402 can be roughly divided into several categories:

If we only look at the high-frequency sellers currently indexed in x402scan, the top is very concentrated. As of my check, the seller with the highest number of transactions is BlockRun. x402scan shows that BlockRun-related origins have generated approximately 13.12 million transactions, about 150,000 USDC in total amount, and about 998 unique buyers, primarily occurring on Base. This scale is far higher than the second place, indicating that in the x402scan metric, high-frequency trading is currently highly concentrated in a few AI routing/payment layer services.

The second tier is closer to the form of "APIs that agents can purchase":

• twit.sh: approximately 80,000 transactions, about 502 USDC, providing real-time data for AI agents on X/Twitter;
• StableEnrich: approximately 50,000 transactions, about 1,889 USDC, providing company/lead/web enrichment APIs;
• Otto AI x402: approximately 31,000 transactions, about 56 USDC, covering market intelligence, DeFi, and trading tools;
• agentutility: approximately 22,000 transactions, about 500 USDC, providing a paid x402 API catalog;
• JarvisClaw API: approximately 13,000 transactions, about 503 USDC;
• OneSource: approximately 9,800 transactions, positioned for AI agents using Ethereum RPC.

If we look at settlement amounts instead of transaction numbers, the rankings change. BlockRun remains first, but dTelecom x402 Gateway generates about 12,600 USDC from approximately 6,108 transactions, indicating that the single prices for real-time communication/voice services like WebRTC, STT, and TTS may be higher. Bitrefill x402 has only about 326 transactions, but the total amount exceeds 1,000 USDC; Laso Finance's crypto prepaid cards/gift cards also generate higher settlement amounts with fewer transactions.

This set of data has two implications.

First, the real use of x402 is not a single scenario. The high-frequency rankings resemble APIs, model routing, data queries, RPC, and social data; the settlement amount rankings begin to show communication, gift cards, prepaid cards, and physical consumption entry points that are closer to payment/commerce scenarios.

Second, transaction numbers and commercial value cannot be equated directly. twit.sh has a high number of transactions but a small settlement amount; dTelecom and Bitrefill have fewer transactions but more significant amounts. For an early protocol like x402, it is essential to look at both frequency and value, rather than just cumulative transactions.

The first category is APIs and data endpoints.

This is the most natural scenario. Developers can set an API, database, model capability, or content resource as a paid endpoint. The caller first receives a 402 Payment Required, and after completing the payment, they gain access.

If in the future Agents need to frequently call real-time data, professional databases, vertical APIs, and MCP tools, this type of scenario will be very important.

The second category is infrastructure platforms like Cloudflare.

Cloudflare has announced the construction of a Monetization Gateway based on x402, aiming to allow websites, APIs, datasets, and MCP tools to complete charging, verification, and access control on the edge network.

The significance of this is not just that "Cloudflare supports stablecoin payments."

More importantly, if payment verification occurs at the infrastructure layer, small teams do not necessarily need to build a complete billing system themselves to charge Agents a few cents or a few dollars.

The third category is wallets and Agent SDKs.

For Agents to make payments, they need wallets. However, these wallets cannot just be an address that holds USDC. They must have budgets, limits, whitelists, approvals, human confirmations, and audit trails.

This is also why the subsequent competition for x402 will not only occur within the protocol itself but will also happen in Agent Wallets, spending policies, identity, observability, and dispute flows.

The fourth category is experimental activities and incentive scenarios.

These scenarios will contribute a high number of transactions but may not necessarily represent long-term commercial demand. Activities like PING are typical examples: they can demonstrate the low cost and high frequency capabilities of x402 but cannot be directly equated with enterprise-level Agent commerce.

This is not a bad thing; early protocols often require experiments, gamified activities, and developer testing to cold start. However, when writing data reviews, it is essential to separate this from real commercial workflows.

Current Data Indicates x402 Has Moved Beyond the "Purely Conceptual" Stage

If we consider x402 as a protocol that will officially launch in May 2025, achieving observable transactions in the range of 100 million to 150 million within about a year is already a rapid advancement.

This at least indicates three things.

First, the direction of HTTP-native payment is not unused.

Developers are willing to try integrating payments into resource access processes rather than always relying on traditional checkouts, subscriptions, and API keys.

Second, there is indeed a demand for small, high-frequency, machine-executable payments.

These demands do not necessarily all come from mature Agents, but they have already proven that when costs are low enough and processes are simple enough, micro-payments between software will naturally occur.

Third, the narrative around x402 is spreading from a Coinbase single-point protocol to Cloudflare, CoinGecko, Circle Wallets, Stripe Machine Payments, Agent SDKs, and various developer experiments.

This indicates that the true potential of x402 is not to become a standalone app but to become an embedded payment primitive. The more successful it is, the less users should "see x402"; users will only see that an Agent automatically purchased data, a model call, a security check, or a tool execution while completing a task. The payment occurred, but it is no longer the main interface.

However, It Has Not Yet Proven That "Agentic Payment Is Mature"

Currently, there are three areas that require caution:

First, the transaction metrics are not transparent enough. We need to know more clearly: which transactions come from real service purchases, which come from testing, which come from incentive activities, and which come from bot behavior.

If a protocol has 100 million transactions, but most of them occur around a short-term activity, then its commercial implications and long-term API payments are entirely different.

Second, the settlement amounts are not large enough. A volume of tens of millions of dollars can prove that x402 has cash flow, but it does not yet prove that it has become mainstream payment infrastructure. A more reasonable judgment now is that x402 has entered the early stage of real use but is still looking for core scenarios with high retention and high repurchase rates.

Third, the issues after payment have not been resolved. The ability for Agents to make payments is just the first step. Once in a commercial environment, businesses and users will ask: who authorized the Agent to spend this money? How much can it spend at most each time? If the service is not delivered, can it be refunded? If the Agent is induced to make a payment through prompt injection, who is responsible? If a resource call fails, can payment and delivery be handled atomically? These questions do not belong to the payment rail itself, but they determine whether the payment rail can be genuinely adopted.

The next phase of Agentic Payment is not just the continued growth of x402 transaction counts. More importantly, it involves spending policies, audit trails, refunds, idempotency, dispute resolution, and human-in-the-loop. In other words: enabling Agents not only to make payments but to be safely allowed to make payments.

The Key Issue of x402 Has Shifted from "Can It Pay" to "Is It Worth Paying Automatically"

The data since the launch of x402 is already sufficient for us to take it seriously. 100 million Base-only transactions, around 150 million cumulative cross-chain transactions, and settlement amounts in the range of 40 to 50 million dollars indicate that the protocol is not just lingering in white papers and demos.

However, these numbers are still not enough for us to jump to the conclusion that "Agentic commerce has exploded". A more accurate statement is: x402 has proven that machines can complete small payments through HTTP-native methods.

It still needs to continue proving that these payments can stably serve real workflows. In the coming year, I will focus on four data points:

First, the number of real paid services, rather than just the number of transactions.

Second, the repeat usage rate, rather than one-time activity peaks.

Third, the average transaction amount and scenario distribution, rather than cumulative volume.

Fourth, the maturity of Agent wallets and spending policies, rather than the attractive narrative of the protocol itself.

If these indicators continue to improve, x402 will not just be a Crypto payment protocol; it may become a foundational protocol when Agents purchase external capabilities.

The significance of this lies not in teaching AI to spend money, but in the fact that when software begins to work for people, it finally needs its own commercial settlement method.

Conclusion

The 150 million transactions of x402 indicate that Agentic Payment has moved beyond the purely conceptual stage, but the real question is not whether the transaction count can continue to grow, but whether these payments can enter real Agent workflows and become controllable in terms of authorization, delivery, refunds, and liability boundaries.