Skip to content
Report · auto-updating

State of the agent economy

Live measurement of the agent web: supply funnel, live-rate by ecosystem, protocol adoption, inflow concentration, and 30-day growth. Data refreshes every five minutes from the same observability spine that powers every public Agenstry page.

As of 2026-09-21 13:59:21 UTC methodology JSON CSV cite this archive
Weekly briefing · 2026-W39 archived view
Sep 21, 2026

The agent web is broadening faster than it is becoming dependable. As of 2026-09-21T13:05:54.820552+00:00, Agenstry indexed 9,653 candidates, up 382 (+4.1%) week on week, while live endpoints rose by 70 (+9.0%) to 846; yet only 118 cards were signed and only 358 payment-labelled services had authoritative runtime payment evidence. The commercial layer is real but highly concentrated: 1,305 observed earning wallets connected to 2,544 services generated $216,986.77 over 30 days, with the top seller accounting for 82.3% of gross volume.

Reach is expanding; operational conformance remains scarce

Agenstry discovery-to-runtime funnelIndexed9.7KValid card7.3KLive846Signed118Earning (30d)1.3K

The principal operating fact remains the gap between declaration and execution. Of 9,653 indexed candidates, 7,337 returned a parseable A2A or MCP card, but 846 were live JSON-RPC or introspectable MCP endpoints. That represents 8.8% of the indexed universe, compared with 76.0% returning a valid card. Only 1.2% were signed. The implication for buyers, registries and enterprise procurers is that discovery metadata should be treated as a lead, not a service-level guarantee.

Source quality differs sharply. The a2aregistry cohort had an 83.3% live rate, registry 63.0%, and a2a_registry_org 40.7%; by contrast, n8n_workflows yielded no live endpoints, while recrawl_cold produced 0.7%. That dispersion supports an operating model in which high-yield registries receive priority refresh and assurance checks, while broad code and workflow discovery remains a prospecting channel. The week’s rise in live endpoints outpaced indexed growth, but the 30-day new-agent count slipped to 2,660 from 2,787, down 127 (-4.6%) week on week.

Payments: x402 dominates listings, while revenue remains concentrated

Declared payment-protocol distribution5.0Kx4024.9K · 96.8%MPP90 · 1.8%A2A x40244 · 0.9%AP223 · 0.5%L4024.00 · 0.1%Stripe SPT1.00 · 0.0%

x402 accounts for 4,862 of the declared payment-protocol records, versus 90 for MPP, 44 for A2A x402 and 23 for AP2. But the evidence base is thin: of 5,024 labelled services, 358—or 7.1%—were self-declared live through an unpaid HTTP 402 response or a served /.well-known/x402 manifest. Agenstry’s runtime challenge remains the meaningful test, rather than a directory listing, static card, or catalogue entry.

The revenue picture reinforces the distinction between listings and functioning markets. There were 214 active sellers and 476 payer-ledger wallets over the measurement window. Median revenue was $0.255 and p95 revenue $44.6397, while the top five sellers captured 93.4% and the top ten 94.9%; the Gini coefficient was 0.993 and HHI 6819.0. The HHI eased from 6,840.90 to 6,819.00, a decline of 21.90 (-0.3%), but not enough to change the structural conclusion: agent-to-agent commerce is presently a narrow set of high-volume settlement relationships, not a distributed long-tail market.

The week-on-week “earning” count fell from 2,502 to 1,305, down 1,197 (-47.8%). Readers should interpret that measure precisely: it is observed inbound USDC at declared x402 payTo addresses, counted one time per wallet; 1,827 of 2,104 scannable wallets were scanned, and the 1,305 wallets link to 2,544 services. It is explicitly not an ecosystem-wide revenue total.

Protocol work is shifting toward validation, statefulness and delegated authority

A2A’s roadmap, updated September 15, prioritises v1.1 robustness, bidirectional streaming, a CLI for coding harnesses, elicitation and multi-turn workflows. It also identifies the A2A Inspector and Technology Compatibility Kit as validation efforts—an important acknowledgement that interoperability depends on testable behavior, not merely compatible cards. A2A’s roadmap (github.com)

MCP is moving from specification release into implementation and professionalisation. The July 28 specification introduced a stateless core, multi-round-trip requests, header-based routing, cacheable list results, authorization hardening and a formal extensions framework; official SDK repositories continued updating through September 21. MCP’s specification announcement and official repositories (github.com) The Agentic AI Foundation’s September 14 MCP Associate certification is a further institutional signal that implementation security and governance are becoming distinct labour-market skills. AAIF announcement (linuxfoundation.org)

Delegated payments are becoming a governance problem, not only a checkout problem

AP2 v0.2 frames autonomous payments around signed Checkout and Payment Mandates, with a trusted surface required to obtain the authorisations and deterministic code required for validation. Its documentation explicitly includes both autonomous and human-present flows, and the specification is now being standardised through FIDO working groups. AP2 documentation and v0.2 specification (github.com)

Research is converging on the residual risk: signatures can preserve transaction integrity after authorisation without proving that the agent’s pre-authorisation context accurately represented the user’s intent. A recent AP2 security analysis identifies threats across transaction phases and deployment architectures; a separate formal study of x402, MPP, ACP and AP2 reaches the related conclusion that delegated authority must remain bound to economic and service outcomes across the lifecycle. AP2 security analysis and formal analysis of agent payment protocols (arxiv.org)

Regulators and payment networks are following the same path. EMVCo has opened feedback on a framework for card-based agentic payments focused on intent, delegation, agent identity and transaction signalling, while the UK government has pointed to ongoing payments-regulation modernisation and FCA sandbox work on agent-led payments and commerce. EMVCo framework consultation and UK parliamentary response (emvco.com)

What to watch next week

The specific item to track is whether A2A’s validation tooling moves from roadmap language into broadly usable compatibility tests and published results. That would address the most consequential gap in this week’s data: the distance between an agent that can be found, an endpoint that answers, and a service that can safely transact.

Sources

  1. A2A/docs/roadmap.md at main · a2aproject/A2A · GitHub
  2. modelcontextprotocol/blog/content/posts/2026-07-28-spec-ga/index.md at main · modelcontextprotocol/modelcontextprotocol · GitHub
  3. Agentic AI Foundation Launches MCPA Certification to Validate MCP Expertise
  4. AP2/docs/index.md at main · google-agentic-commerce/AP2 · GitHub
  5. Beyond the Mandate: A Systematic Security Analysis of the Agent Payments Protocol (AP2)
  6. EMVCo Requests Feedback on Framework for Secure, Interoperable and Scalable Card-Based Agentic Payments | EMVCo
  7. A2A’s roadmap
  8. MCP’s specification announcement
  9. official repositories
  10. AAIF announcement
  11. AP2 documentation
  12. v0.2 specification
  13. AP2 security analysis
  14. formal analysis of agent payment protocols
  15. EMVCo framework consultation
  16. UK parliamentary response
1 · Supply funnel

Most of the agent economy is indexed, not yet operational.

Counting agents conflates supply with operation. Below is the corpus broken into five stages: what fraction of indexed agents pass each gate. The decay curve is the honest read on how mature the agent web is at any given moment.

Stage Count % of indexed Definition
Indexed 9,653
100.0%
Every candidate we have seen across every source.
Valid card 7,337
76.0%
Returned a parseable A2A/MCP card on probe.
Live 845
8.8%
Live JSON-RPC / introspectable MCP.
Signed 118
1.2%
Card published with a JWS signature.
Earning (30d) 1,305
13.5%
Settlement wallets where we observed inbound USDC to the x402 payTo address declared by an indexed agent, one count per wallet (2544 linked services). Measured over the wallets scanned so far (1827 of 2104 with a wallet) — not an ecosystem-wide total.
2 · Live-rate by ecosystem

Which source registries actually ship live agents.

Each row shows one source registry (Anthropic's MCP directory, Smithery, Glama, Postman, GitHub well-knowns, Wayback CDX scans) alongside how many of the agents it lists actually respond live on probe. The gap between seen-count and live-count is the source's effective freshness.

Source Seen Live Live rate
recrawl_hot 3,932 572 14.5%
agentic_market 3,728 95 2.5%
recrawl_warm 3,167 197 6.2%
cdp_discovery 2,265 65 2.9%
manifests 1,817 389 21.4%
mcp_registry 1,747 277 15.9%
n8n_workflows 972 0 0.0%
github_code 896 90 10.0%
x402_list 719 47 6.5%
registry 557 350 62.8%
lists 531 46 8.7%
recrawl_cold 290 2 0.7%
a2aregistry 275 228 82.9%
a2a_registry_org 209 84 40.2%
github_topics 163 7 4.3%
3a · By protocol kind

A2A vs MCP vs paid_api

paid_api
4,766
a2a_agent
2,571
workflow
972
3b · By observed payment-protocol label

Index composition, not global market share. x402 has multiple dedicated catalogs and a live HTTP probe; AP2, L402 and Stripe ACP are predominantly visible only when cards declare them.

x402 / AP2 / Stripe MPP / L402

x402
4,862
mpp
90
a2a_x402
44
ap2
23
l402
4
stripe_spt
1
3b · Evidence grade

How we know they take payment

358 of 5,024 payment-labelled agents (7.1%) proved it themselves — their own endpoint answered an unpaid request with an HTTP 402 challenge, or the host serves a /.well-known/x402 manifest. The rest is declared in a static agent card or described by a third-party catalog.

Facilitator feed
3,115
Catalog listing
1,041
Agent card
402
Live 402 challenge
259
/.well-known/x402
89
Official protocol directory
88
Not re-probed yet
20
mpp_challenge
9
mpp_manifest
1
3c · Indexed protocol labels over time · last 12 weeks

Newly indexed agents per week, by payment label

Changes reflect both ecosystem activity and changes in crawler source coverage.

2026-W27 · a2a_x402: 4 2026-W27 · x402: 10 27 2026-W28 · a2a_x402: 5 2026-W28 · ap2: 2 2026-W28 · l402: 1 2026-W28 · mpp: 1 2026-W28 · x402: 1137 2026-W29 · a2a_x402: 2 2026-W29 · ap2: 1 2026-W29 · mpp: 1 2026-W29 · x402: 244 29 2026-W30 · a2a_x402: 2 2026-W30 · mpp: 4 2026-W30 · x402: 296 2026-W31 · a2a_x402: 2 2026-W31 · ap2: 1 2026-W31 · mpp: 81 2026-W31 · x402: 263 31 2026-W32 · a2a_x402: 10 2026-W32 · ap2: 2 2026-W32 · x402: 219 2026-W33 · a2a_x402: 4 2026-W33 · ap2: 3 2026-W33 · x402: 224 33 2026-W34 · a2a_x402: 1 2026-W34 · ap2: 1 2026-W34 · x402: 260 2026-W35 · ap2: 1 2026-W35 · mpp: 2 2026-W35 · x402: 244 35 2026-W36 · a2a_x402: 4 2026-W36 · x402: 403 2026-W37 · a2a_x402: 2 2026-W37 · ap2: 1 2026-W37 · x402: 344 37 2026-W38 · a2a_x402: 1 2026-W38 · x402: 261 2026-W39 · x402: 46 39
a2a_x402 ap2 l402 mpp x402
4 · Inflow concentration · 30d

A few settlement wallets receive most of the money.

Across 1,305 earning wallets backing 2,544 linked services, the distribution is heavily power-law. Wallet-level Gini and HHI avoid multiplying transfers when several catalog entries share one recipient.

Observed inflow · 30d
$216986.77
Recipient wallets
1,305
Active sellers · 30d
214
Median · wallet
$0.255
p95 · wallet
$44.64
Top 1 share
82.3%
Top 5 share
93.4%
Gini
0.993
extreme inequality
HHI
6819
highly concentrated
5 · 30-day growth

New agents arriving daily, by first-seen.

Daily count of agents we discovered for the first time. Sourced from every registry + open-web crawl we run.

2026-08-23 peak 652/day 2026-09-21
New agents · 30d
+2,660
Earning agents · 30d
443
observed in the snapshot table

Cite this report

Public domain numbers (CC BY 4.0). Pick the format that fits. The page auto-updates so the URL is the canonical pointer; the date stamps the snapshot you cited.

Plain text
Agenstry Research. "State of the Agent Economy." Accessed 2026-09-21. https://agenstry.com/reports/state-of-agent-economy
BibTeX
@misc{ '{' }}agenstry_state_of_agent_economy,
  title  = {State of the Agent Economy},
  author = {Agenstry Research},
  url    = {https://agenstry.com/reports/state-of-agent-economy},
  note   = {Accessed 2026-09-21},
  year   = {2026}
}
Embed (iframe)
<iframe src="https://agenstry.com/reports/state-of-agent-economy" width="100%" height="900" frameborder="0"></iframe>
Scope What this report measures, and what it doesn't

All inflow numbers are on-chain: direct eth_getLogs scans of USDC Transfer events (plus EURC on Base + Ethereum) into each indexed agent's payment_wallet across Base, Ethereum, Polygon, Arbitrum, Optimism (EVM via eth_getLogs) and Solana (SPL via getSignaturesForAddress). Reproducible from public RPC; not based on self-reporting. We report money arriving at those wallets — we do not verify that each transfer was an arm's-length sale, and we do not net out transfers an operator sends itself. See the measurement boundary.

Not included: money agents take via Stripe (per-agent Stripe accounts are private), Patreon / Sponsors, direct credit cards, PayPal, or any off-chain rail. Agenstry's own platform-skill revenue (compose / agent_stats / etc.) is also excluded: it lives in a separate accounting table that never feeds public totals. This is "the observable on-chain slice of the agent economy", not "all agent revenue ever". Off-chain rails will appear here only when the operator opts in to a future verified-reporting feed.

Methodology

Every figure on this page is computed from the same observability tables that power the rest of the Agenstry surface, with no synthetic data and no manual curation. The crawl ingests eight federated sources plus open-web well-known probes. Agents are scored against a 9-criterion conformance methodology. Observed inflow is derived from on-chain USDC/EURC transfer scans across six chains: Base, Ethereum, Polygon, Arbitrum, Optimism, and Solana; AP2 / Stripe MPP / L402 receipts are detected on agent cards but indexed only when the rail is publicly verifiable. The full crawl + scoring pipeline is open behind the federation feed. If our numbers disagree with yours, one of us has a bug, and we would like to know.