> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tythe.finance/llms.txt
> Use this file to discover all available pages before exploring further.

# Market Analysis

> Where agents transact today, what existing layers answer, and what they leave open.

This page sets out the market Tythe enters, using the strongest available data and stating where that data is contested. Figures are dated. Every claim carries a reference. Where sources disagree, the disagreement is shown rather than resolved in Tythe's favour.

### Summary

Stablecoins have become the settlement layer for programmable payments, but only a small fraction of headline volume is real economic payment, and most of that fraction is business-to-business. Agents have begun transacting on open rails, but verified agent-driven commerce is still small, and the analysts who measured it identify the same cause: agents lack verifiable identity, reputation, and bounded authority, so the rail outruns the trust that would let real money flow on it. Every existing layer answers how an agent pays or who it is. None underwrites the principal behind the agent, underwrites the agent, sizes the authority it has earned, enforces that authority at the source of funds, or extends credit against it. That is the layer Tythe builds.

### 1. Stablecoins as the settlement layer

#### **1.1 Gross volume, and what it measures**

Reported stablecoin transfer volume for 2025 ranges from roughly \$28 trillion to more than \$62 trillion depending on inclusion rules: the Bank for International Settlements estimates about \$28 trillion, Artemis Analytics about \$33 trillion (a 72% year-over-year increase), and BCG with Allium more than \$62 trillion \[1]\[2]. Visa's adjusted figure, which strips automated and internal transfers, is about \$10.2 trillion over a trailing twelve months \[3].

These figures measure transfers, not payments. McKinsey, working with Artemis Analytics on December 2025 data, estimates that genuine end-user payments (supplier invoices, remittances, payroll, card spend) run at about 390 billion annualized, roughly 1% of headline volume \[4]. Independent estimates of real-economy payments fall in a 350-550 billion range \[1].

#### **1.2 The payment layer is growing fast, and it is B2B**

Within the real-payment fraction, business-to-business flows dominate: about \$226 billion, or 58% of genuine stablecoin payments, per the McKinsey/Artemis analysis \[4]\[5]. Monthly real payment volume rose from about \$5 billion in January 2024 to more than \$30 billion by early 2026, with the steepest acceleration in the second half of 2025 \[5]. Consumer usage remains negligible by comparison \[5].

McKinsey's interpretation is that stablecoins are emerging as a programmable settlement layer for machines, finance departments, and institutions, with consumer adoption as a secondary effect \[4]\[5].

#### **1.3 Supply and policy**

The US Treasury Secretary has projected stablecoin supply of \$3 trillion by 2030; leading financial institutions project \$2 to 4 trillion over the same period \[4]. The GENIUS Act, in force since July 2025, established a federal framework for payment stablecoins; the IMF estimates its passage removed about \$300 billion, roughly 18%, from the market value of incumbent payment firms, with cross-border firms most affected \[1]. USDC led by annual transaction volume in 2025 at \$18.3 trillion against USDT's \$13.3 trillion \[1].

### 2. Agent transactions, as they actually are

#### 2.1 Headline activity on x402

x402, launched by Coinbase in May 2025 and now governed by the x402 Foundation under the Linux Foundation with backers including Cloudflare, Stripe, AWS, Google, Visa, and Circle \[6], is the most-used agent payment protocol. Coinbase reported 69,000 active agents, 165 million transactions, and about \$50 million cumulative volume by late April 2026 \[7]. An independent multi-chain tracker counted 157.4 million settled transactions and \$41.1 million in stablecoin volume across seven chains and eighteen facilitators as of 19 July 2026 \[8]. The protocol's own dashboard showed about 75 million transactions and \$24 million in volume for the trailing thirty days in mid-July 2026 \[9].

#### 2.2 What survives screening

The headline figures overstate agent commerce. TRM Labs screened the \$52.7 million settled through x402 since May 2025, removing self-payments, bulk flows concentrated in one or two payers, and sellers with fewer than ten distinct buyers; about half the value fell away, leaving \$25.6 million of plausibly genuine buyer-seller commerce. Of that, TRM estimates 0.6% (strict model) to 7.5% (permissive model) is plausibly agent-driven, a 2026 run-rate of roughly \$5,000 to \$11,000 per month in verified agentic dollar volume \[10]. CoinDesk reporting cited by a separate tracker puts real daily commerce on the order of \$28,000, with roughly half of total activity classified as gamified or test traffic \[6]. Transactions of \$1 or more grew from 49% of volume in early 2025 to 95% by early 2026, so the dollar volume that exists is shifting toward larger, less agent-typical payments \[11]. Daily transaction counts also fell about 92% from a December 2025 peak of 731,000 to 57,000 in February 2026 \[12].

TRM's conclusion is the one this analysis rests on: "Until verifiable agent identity and reputation exist, the rail will keep outrunning the wallets on it" \[10].

#### 2.3 Identity and the other rails

ERC-8004, the agent identity, reputation, and validation standard, went live on Ethereum mainnet in January 2026 with backing from the Ethereum Foundation, Coinbase, Google, and MetaMask; about 24,000 agents had registered by May 2026 \[12]. Google's Agent Payments Protocol launched with more than sixty organisations including PayPal, Coinbase, Mastercard, and American Express \[12]. Academic security researchers documented replay and cross-resource free-riding flaws in production x402 facilitator implementations as of mid-2026 \[11].

#### 2.4 The honest read

The rail works and its backers are serious. Standard adoption is strong; commercial adoption is not yet \[6]. The gap is not throughput or cost. It is that no one can yet answer, for a given agent, who stands behind it, what it may do, and whether it has earned that, in a form a counterparty or a lender can verify and enforce.

### 3. The layers that exist

| Layer                            | Examples                                                                               | Answers                                                         | Leaves open                                                                 |
| -------------------------------- | -------------------------------------------------------------------------------------- | --------------------------------------------------------------- | --------------------------------------------------------------------------- |
| Payment rails                    | x402 \[6], Stripe agent payments, Google AP2 \[12], Visa and Mastercard agent programs | How an agent pays                                               | Whether it should be allowed to                                             |
| Agent wallets and secured credit | Floe, Skyfire, Bank of Bots                                                            | Where an agent holds funds; identity; a collateral-secured line | Underwriting the principal and the agent; authority that adjusts to conduct |
| Policy engines                   | Newton Protocol; draft ERCs 8196 and 8199                                              | Whether an action passes a rule written by a human              | Capital behind the rule; credit; learning from conduct                      |
| Identity registries              | ERC-8004 \[12]                                                                         | Who an agent is; what others report about it                    | What it may do with money, and how much                                     |

In every layer the key is the authority. A leading payments provider's own agent documentation states that there is no per-agent token concept at the network level and that scoping is the developer's responsibility at the key-management layer. A leaked key is a drained wallet; that is the ceiling on how much money anyone will let an agent touch.

### 4. The unsolved problem

Trust in cross-system finance fails at the verification layer, not the settlement layer. In April 2026 a forged cross-chain message minted about 116,500 unbacked tokens that were deposited as collateral into Aave and borrowed against for roughly \$190 million across two chains within minutes; Aave's contracts, oracles, and liquidation engine performed as designed, and the loss was inherited from trusting an unverified external claim \[13]. The same structure recurs in agent finance whenever authority is inferred from key possession rather than underwritten and enforced at the source of funds.

The problem, stated precisely, has four parts, and each existing layer solves at most one:

1. **Who stands behind the agent.** A principal with legal standing, an underwritable balance sheet, and recourse.
2. **What the agent may do.** Bounded, expiring, revocable authority, enforced where the money is, not at an API gateway.
3. **Whether it has earned that.** Underwriting of the agent on its own conduct, so authority is sized from evidence and narrows when conduct changes.
4. **Credit against that authority.** Working capital extended to the principal and drawable by agents within limits, with loss absorption ordered and disclosed.

### 5. Sizing the opportunity

Two lenses, both from sources outside the crypto industry.

* **Agentic commerce.** McKinsey projects a global agentic commerce opportunity of \$3 to 5 trillion by 2030 \[14]. Gartner projects that machine customers could account for up to 20% of revenue by 2030 \[14].
* **B2B stablecoin payments.** Genuine B2B stablecoin payments were about \$226 billion in 2025 and growing sixfold in under two years at the monthly level \[4]\[5]. Businesses that have used stablecoins report cost savings of 10% or more in 41% of cases, mostly on cross-border payments \[1].

Tythe's addressable market is the intersection: principals whose agents will transact on stablecoin rails and who need underwritten, enforceable authority and credit to let them. Today that intersection is measured in thousands of dollars per month of verified agent commerce \[10]. The thesis is that the intersection is small because the authority layer is missing, not because the demand is.

### 6. Where Tythe fits

Tythe uses every layer in Section 3 and competes with none. Agents pay on x402. Identities live in ERC-8004, where Tythe also publishes bounded standing. Screening comes from established providers. Venues are curated, not built.

Tythe adds the delegated-authority layer: a verified, underwritten principal that owns a noncustodial account; agents bound to it by signed authorization and underwritten on conduct; mandates enforced at the principal's own address; credit extended against the principal's rating and drawable by agents within their share; and an intelligence layer that improves with every transaction.

| Compared with         | They provide                                                                                                    | Tythe differs in                                                                                                                                                       |
| --------------------- | --------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Floe                  | Agent wallet, fiat ramps, collateral-secured working-capital lines, x402, framework adapters, ERC-8004 identity | Authority is underwritten and adjusts to conduct; credit is underwritten rather than collateral-secured; the principal, not the agent, is the account and the borrower |
| Newton Protocol       | Verifiable policy evaluation by a decentralised operator network                                                | Policy enforced deterministically at the principal's account, bound to real capital and credit, derived from underwriting, with no operator set and no token           |
| Skyfire, Bank of Bots | Agent-centric identity, trust credentials, credit history for agents                                            | Principal-centric: the party with legal standing, recourse, and an underwritable balance sheet is the unit of underwriting; agents are underwritten as its actuators   |

### 7. Limitations of this analysis

Agent-payment data is young, vendor-reported, and inconsistent across trackers; figures should be read as point-in-time snapshots with the stated dates. Stablecoin volume estimates vary by an order of magnitude with methodology; this page prefers screened and adjusted figures where they exist. Forward projections from McKinsey and Gartner are estimates, not measurements. Competitor descriptions are drawn from public materials as of September 2026 and may lag product changes.

### References

1. Reap, "Stablecoin Statistics and Data 2026", August 2026, citing BIS, Artemis, BCG and Allium, CoinDesk Data, McKinsey, and IMF. [https://reap.global/blog/stablecoin-statistics-2026](https://reap.global/blog/stablecoin-statistics-2026)
2. Plasma, "Stablecoin Transaction Volume Trends in 2026", February 2026, citing Artemis/Bloomberg. [https://www.plasma.to/learn/stablecoin-transaction-volume](https://www.plasma.to/learn/stablecoin-transaction-volume)
3. Visa, adjusted stablecoin volume (trailing twelve months), as cited in \[2].
4. McKinsey & Company with Artemis Analytics, "Stablecoins in payments: What the raw transaction numbers miss", February 2026. [https://www.mckinsey.com/industries/financial-services/our-insights/stablecoins-in-payments-what-the-raw-transaction-numbers-miss](https://www.mckinsey.com/industries/financial-services/our-insights/stablecoins-in-payments-what-the-raw-transaction-numbers-miss)
5. Stablecoin Insider, "The Institutional Gap in Stablecoins: B2B Payments vs End User Volume", March 2026, citing McKinsey/Artemis and BVNK. [https://stablecoininsider.org/the-institutional-gap-in-stablecoins/](https://stablecoininsider.org/the-institutional-gap-in-stablecoins/)
6. Presenc AI, "x402 Protocol Adoption Tracker 2026", snapshot 15 May 2026, citing CoinDesk. [https://presenc.ai/research/x402-protocol-adoption-tracker-2026](https://presenc.ai/research/x402-protocol-adoption-tracker-2026)
7. Eco, "x402 Protocol Explained", August 2026, citing Coinbase (late April 2026). [https://eco.com/support/en/articles/14839402-x402-protocol-explained](https://eco.com/support/en/articles/14839402-x402-protocol-explained)
8. agenteconomy.to, "How many x402 transactions have been processed?", data as of 19 July 2026. [https://agenteconomy.to/stats/x402-transactions](https://agenteconomy.to/stats/x402-transactions)
9. Concordium, "x402 Explained: The HTTP Payment Protocol for AI Agents", September 2026, citing x402.org dashboard (mid-July 2026). [https://www.concordium.com/article/x402-explained-agentic-payments-identity](https://www.concordium.com/article/x402-explained-agentic-payments-identity)
10. TRM Labs analysis of x402 agentic volume, as reported by Spoted Crypto, "TRM Labs x402 Report: AI Agent Crypto Payments Analyzed (2026)", September 2026. [https://www.spotedcrypto.com/trm-labs-x402-ai-agent-payments-report-2026/](https://www.spotedcrypto.com/trm-labs-x402-ai-agent-payments-report-2026/)
11. Crypto AI Meta, "Understanding the x402 Protocol for High-Frequency Microtransactions", August 2026. [https://cryptoaimeta.com/x402-protocol-microtransactions/](https://cryptoaimeta.com/x402-protocol-microtransactions/)
12. Nevermined, "40 Stablecoin Payments for AI Agents Statistics", May 2026, citing Solana, BlockEden, Cambrian Network, and Google. [https://nevermined.ai/blog/stablecoin-payments-ai-agents-statistics](https://nevermined.ai/blog/stablecoin-payments-ai-agents-statistics)
13. Public reporting on the April 2026 Kelp DAO rsETH bridge exploit and its effect on Aave markets (a forged cross-chain message; approximately \$190 million borrowed against unbacked collateral). Primary incident post-mortems from the affected protocols.
14. McKinsey & Company, agentic commerce projection (\$3 to 5 trillion by 2030), and Gartner, machine-customer revenue projection (up to 20% by 2030), as cited in Nevermined, "47 Agentic Economy Transaction Volume Statistics", May 2026. [https://nevermined.ai/blog/agentic-economy-transaction-volume-statistics](https://nevermined.ai/blog/agentic-economy-transaction-volume-statistics)

Figures accessed September 2026. Where a figure is cited through a secondary source, the primary is named in the reference and should be consulted directly for diligence.