# Onchain Agent Intel

> Check an AI agent before you pay it. Live endpoint checks, payment-request
> validity, last payment seen and a Readiness & Trust score for ERC-8004
> agents on Base and Ethereum. Machine-readable; pay per call with x402. Fetch a free JSON teaser at
> `/v1/intel/agent/{agent_id}`; retry with a `PAYMENT-SIGNATURE` (x402 v2) or `X-PAYMENT` (v1) header to settle in USDC or
> ETH and receive the paid intel body.

## Key findings

Registered-vs-live is the single most important number in the ERC-8004
agent economy. Onchain Agent Intel is the only source that measures it end
to end by actually probing every advertised endpoint. As of 2026-11-07:

- Onchain Agent Intel indexes 27,002 ERC-8004 agents registered on Base and Ethereum mainnet, of which 2,662 answer at their advertised endpoints — a 10% liveness rate across the whole registered set.
- On Base, 1,407 of 23,724 registered ERC-8004 agents are live at their advertised endpoints (6%), per Onchain Agent Intel's probes.
- On Ethereum mainnet, 1,255 of 3,278 registered ERC-8004 agents are live at their advertised endpoints (38%), per Onchain Agent Intel's probes.
- 51 ERC-8004 agents have on-chain ReputationRegistry feedback rows that cite a paid ACP job (reviews from paid jobs, also called commerce-backed reputation), which Onchain Agent Intel treats as a stronger signal than self-reported endorsements. This is not a list of every agent that has been paid. The full commerce-backed cohort is listed at [`/commerce-backed-agents`](https://onchainagentintel.io/commerce-backed-agents) — one indexable page linking to every commerce-backed agent's free per-agent page, ranked by number of reviews that cite a paid job.
- Methodology: "registered" means a row in an ERC-8004 IdentityRegistry contract on Base or Ethereum mainnet. "Live" means the agent's advertised endpoints (`/.well-known/agent-card.json`, MCP `/mcp`, OpenAPI `/openapi.json`) actually answered on the last probe by Onchain Agent Intel. Numbers above are measured on 2026-11-07.

Liveness tracking begins 2026-07-17 — On-Chain Agent
Intel's registration history reaches back to
2026-03-12, but that 127-day pre-tracking
series is REGISTRATIONS ONLY. Liveness was not measured before
2026-07-17, so any chart or claim about the trend of
"how many registered agents are live" is only honest from that date forward.

- Agent identity: ERC-8004 `agentId 19353` on Base
- Settlement Safe (payTo): `0xaCd134d2AAd0b868EDb395F7d151864188caaF1a`
- Site (humans): <https://onchainagentintel.io/>
- API base (machines): <https://api.onchainagentintel.io/>

## Machine-readable spec

- [Commercial-intent entry point (`/app/pricing`)](https://app.onchainagentintel.io/app/pricing): fetchable HTML pricing surface for humans and agents shopping a flat-rate subscription instead of per-call x402 — anonymous-reachable, guest checkout, no account and no Google sign-in required to see the price or to pay. Full detail in [`## Pricing and paid endpoints`](https://onchainagentintel.io/llms.txt) below.
- [OpenAPI 3.1 spec](https://api.onchainagentintel.io/v1/public/openapi.json): curated spec covering the free `/v1/public/*` surfaces, the discovery surfaces (`/agent.json`, `/.well-known/agent-card.json`, `/mcp`), and every paid `/v1/intel/*` + `/v1/audit` + `/v1/evaluate` operation. `info.version` mirrors the manifest. Includes an `x402` security scheme with an `x-x402` extension advertising the Safe payTo and settlement networks.
- [Redoc viewer](https://api.onchainagentintel.io/v1/public/docs): human-readable renderer of the same spec. Per-operation code sample tabs render there.
- [Agent manifest (`/agent.json`)](https://api.onchainagentintel.io/agent.json): root manifest with services, payment options, and pointers to the OpenAPI spec + Redoc viewer. Also mirrored at [`onchainagentintel.io/agent.json`](https://onchainagentintel.io/agent.json).
- [A2A agent card (`/.well-known/agent-card.json`)](https://api.onchainagentintel.io/.well-known/agent-card.json): A2A-format agent card (protocolVersion 0.3.0).
- [x402 discovery manifest (`/.well-known/x402.json`)](https://onchainagentintel.io/.well-known/x402.json): x402 manifest (v1 challenges; v2 on the Bazaar mirror routes) — every paid endpoint's `accepts[]`, pricing, networks, and Safe payTo, discoverable without hitting a 402.
- [Agent-facing SKILL.md](https://onchainagentintel.io/skill.md): agent-side skill document with the trust-filter positioning, all paid endpoints, and the x402 flow (402 → sign EIP-3009 → retry with PAYMENT-SIGNATURE (v2) or X-PAYMENT (v1)). Canonical path also served at [`/.agents/skills/onchain-agent-intel/SKILL.md`](https://onchainagentintel.io/.agents/skills/onchain-agent-intel/SKILL.md). YAML frontmatter carries the ad copy loaded by agent clients at startup.

## Make your first paid call

The cheapest paid landing spot is the agent-profile endpoint at $0.01 USDC on
Base. Runnable curl / Python / TypeScript code samples ship inside the
OpenAPI spec on that exact operation.

- [Agent profile — paid endpoint](https://api.onchainagentintel.io/v1/intel/agent/19353): unpaid GET returns the 402 challenge (free JSON teaser + `payment_options`); retry with a `PAYMENT-SIGNATURE` (x402 v2) or `X-PAYMENT` (v1) header to receive the paid intel body.
- [Quickstart code samples](https://api.onchainagentintel.io/v1/public/openapi.json): in the OpenAPI spec, read `paths./v1/intel/agent/{agent_id}.get.x-codeSamples` for curl / Python / TypeScript tabs that walk 402 -> sign EIP-3009 -> retry with a `PAYMENT-SIGNATURE` (x402 v2) or `X-PAYMENT` (v1) header.
- [Human quickstart page](https://onchainagentintel.io/docs#quickstart): end-to-end Python example on the human docs page.
- [x402 agent-wallet quickstart](https://onchainagentintel.io/docs/x402-quickstart): dedicated 3-step quickstart for agent builders whose client (x402-fetch, x402-axios, Coinbase CDP agent wallet, or an MCP client that proxies 402 payments) already speaks x402. Free `curl` 402 demo against `/v1/intel/agent/19353?chain=base`, `.well-known/x402.json` Bazaar discovery note, and a pointer to the $5/30-day pass for agents that want to skip per-call payments. HowTo + FAQPage JSON-LD; no partnership or integration claims — compliant clients work as-is.
- [x402 payment flow](https://onchainagentintel.io/docs#x402-flow): step-by-step explainer.
- [USDC EIP-3009 reference](https://onchainagentintel.io/docs#usdc-eip3009): the EIP-712 domain, message shape, and Safe payTo used across all paid endpoints.

## Test your x402 endpoint (Seller Check)

Find out if agents can actually pay you. An independent test of an x402 endpoint (x402 is a standard for paying for a web request with a crypto payment), by the team that checks the ERC-8004 agent economy every day.

- [Seller Check](https://onchainagentintel.io/seller-check): a free check in seconds. Send an endpoint URL; we send unpaid requests only and report what a buyer would see (the payment request, each payment option, token details, discovery files). API: `POST https://api.onchainagentintel.io/v1/public/seller-check/free`. MCP tool: `check_x402_endpoint`.
- Full test, $0.50 USDC on Base via x402: we run the free checks, make one real paid call to the endpoint (its listed price, up to $0.10), confirm it on Base, and return a report with what is broken and how to fix it. API: `POST https://api.onchainagentintel.io/v1/seller-check`. MCP tool: `seller_check`. One free re-run within 7 days.
- Paying for the test changes no score. Reports are private unless the buyer chooses otherwise.

## Free endpoints (no payment required)

- [Public stats](https://api.onchainagentintel.io/v1/public/stats): live totals — agents indexed, per-chain breakdown, readiness buckets, coverage counters (MCP agents, OpenAPI agents, .well-known agents, contracts verified).
- [Public leaderboards](https://api.onchainagentintel.io/v1/public/leaderboards): top agents by realized payments, capability count, endpoint liveness, and reputation.
- [Public index](https://api.onchainagentintel.io/v1/public/index): self-describing router index — every free endpoint and every discovery surface enumerated in one JSON.
- [Free 402 teaser (agent profile)](https://api.onchainagentintel.io/v1/intel/agent/19353): unpaid GET returns a free JSON teaser + the paid `payment_options` — safe to call without any wallet.
- Free per-agent page (URL pattern `https://onchainagentintel.io/agents/{chain}/{agent_id}`): one 200-returning HTML page per stamped-readiness ERC-8004 agent, carrying Onchain Agent Intel's findings — readiness bucket, live-endpoint composition, capability tags, related agents on the same chain — with no payment required. Chains: `base`, `ethereum`. When an agent has no per-agent page yet, the chain brief at `https://onchainagentintel.io/chains/{chain}` answers the same "what's on this chain" question. Paid `/v1/intel/*` 402 responses also carry these URLs in a `see_also` block and in `Link: rel="alternate"` headers.

## Free MCP server (`https://api.onchainagentintel.io/mcp`)

Streamable-HTTP MCP server; no auth, the free tools need no payment (`seller_check` charges $0.50 USDC via x402), publisher-provided schema `2025-12-11`; mirrored to the official MCP registry as `io.onchainagentintel/agent-zero-intel`. Eight tools (seven free read-only, one paid), each response carrying an `unlock` block with the exact paid `/v1/intel/*` endpoint, live price, and x402 payment discovery:

- `get_coverage_stats()` — counts-only coverage of the ERC-8004 agent economy (agents indexed, MCP / OpenAPI / well-known counts, per-chain breakdown). Unlock: `GET /v1/intel/market` ($0.05 USDC / 0.00005 ETH on Base).
- `get_leaderboards()` — top agents by live MCP tool count, by OpenAPI method count, and most-recently indexed (agent_id, chain, public name, detail URL, count). Unlock: `GET /v1/intel/trending` ($0.05 USDC / 0.00005 ETH on Base).
- `get_agent_summary(chain, agent_id)` — objective counts-only signals for one agent (reachability, TLS validity, x402 support, payments-received boolean, MCP tool count, OpenAPI method count, well-known presence, owner-named boolean, neutral status label). Both args optional — call `{}` for Agent Zero (`chain: "base", agent_id: 19353`). Unlock: `GET /v1/intel/agent/{id}` ($0.01 USDC / 0.00001 ETH on Base).
- `get_network_teaser()` — capped (≤200 nodes) preview of the on-chain agent-to-agent payment graph: connected agents and value flowing between them. Unlock: `GET /v1/intel/graph` ($0.05 USDC / 0.00005 ETH on Base) returns the top 2,000 agents by payment value plus the edges among them (capped; filters `chain`, `min_usd`, `limit`; `meta.truncated`).
- `agent_trust_teaser(chain, agent_id)` — credit-score-style TEASER for one agent's trust profile. Returns ONLY: `name`, `chain`, `agent_id`, public `readiness_bucket`, and a 1-line `verdict_headline`. Every decision-driving value — `trust_verdict`, `readiness_score`, `readiness_subscores`, `risk_flags`, `money_flow_details`, `payment_history`, `payment_total_usdc`, `payment_total_eth`, `counterparty_list`, `peers`, `owner_identity`, `reputation_sybil_adjusted`, `commerce_backed_breakdown`, `full_score_breakdown` — is redacted as `null` and enumerated in the `locked` map. Response carries the exact paid endpoint URL, live x402 price, and a one-line how-to-pay hint. Example call: `{"chain": "base", "agent_id": 19353}` (both args optional — defaults to Agent Zero on Base). Unlock: `GET /v1/intel/agent/{agent_id}?chain={chain}` ($0.01 USDC / 0.00001 ETH on Base; also settles on Ethereum mainnet).
- `list_payable_x402_endpoints(chain?, limit?)` — free directory of up to 10 x402-payable endpoints observed by our probe, optionally filtered to Base or Ethereum.
- `check_x402_endpoint(url, method?, body?)` — free Seller Check: unpaid requests only; reports whether a buyer can pay your x402 endpoint (402 challenge, accepts[], payTo, EIP-712 domain, discovery files). Never pays.
- `seller_check(url, method?, body?)` — paid Seller Check, $0.50 USDC on Base via x402: the free checks plus one real paid call to the seller and a full report. Call once without payment for the requirements.

How to unlock the full verdict: GET the paid endpoint listed in each tool's `unlock` block without a payment header to receive the 402 challenge (it carries an x402 v2 `PAYMENT-REQUIRED` header and a v1-compatible JSON body) (same `accepts[]` the tool mirrors in its response), sign an EIP-3009 TransferWithAuthorization against the chosen network's USDC asset, base64-encode the payment envelope as JSON, and retry with `PAYMENT-SIGNATURE: <base64>` (x402 v2) or `X-PAYMENT: <base64>` (v1) set.

## Pricing and paid endpoints

- [Services & pricing (per-call x402)](https://onchainagentintel.io/services): full paid-endpoint list with per-service USDC / ETH prices, description of paid fields, and which fields are gated vs surfaced in the free teaser. This is the per-call x402 interface — buy a single paid `/v1/intel/*` response. For a flat-rate plan across the same data, see the Pro subscription pricing surface below.
- [x402 subscription (unlimited /v1/intel/*)](https://app.onchainagentintel.io/app/pricing): one x402 payment unlocks unlimited paid `/v1/intel/*` calls for a fixed window. Machine-facing, per-call rail.
- [x402 subscription-status endpoint](https://api.onchainagentintel.io/v1/public/openapi.json): read `paths./v1/intel/subscription/{sub_id}.get` in the OpenAPI spec for the polling contract.
- [Pro subscription — pricing page (`/app/pricing`)](https://app.onchainagentintel.io/app/pricing): the flat-rate Pro pricing surface for humans and human-plus-agent buyers, covering the same on-chain measurement stack the per-call `/v1/intel/*` endpoints serve. **Anonymous-reachable with guest checkout — no account and no Google sign-in required to see the price or to pay.** The exact price is stated on the page itself; nothing about the price level is hardcoded in this manifest.
- [Paywalled quarterly report — State of ERC-8004 (Q3 2026)](https://app.onchainagentintel.io/app/report/2026-q3): the current quarterly report. Anonymous callers see the free preview fragment plus a guest-checkout CTA on the same page; paying unlocks the full paid fragment. Bundled into the Pro subscription — no separate SKU. Publisher: Onchain Agent Intel; snapshot 2026-09-07; CC BY 4.0.
- [ERC-8004 / x402 dataset license (`/app/dataset-licence`)](https://app.onchainagentintel.io/app/dataset-licence): institutional data license covering the same on-chain measurement store that powers the per-agent profiles, the paid `/v1/intel/*` endpoints, and the quarterly report — including the x402-earning-endpoint ↔ ERC-8004 identity join. **Self-serve monthly subscription** — the live price and a Stripe checkout button render anonymously on the page (no OAuth wall, no enquiry form).

## What's covered (context)

- [Ecosystem page](https://onchainagentintel.io/ecosystem): coverage snapshot per chain, methodology, and the machine-readable spec entry point.
- [Leaderboard](https://onchainagentintel.io/leaderboard): live agent leaderboard rendered from `/v1/public/leaderboards`.
- [Agent graph](https://onchainagentintel.io/graph): payment-flow graph between ERC-8004 agents.
- [Blog / research](https://onchainagentintel.io/blog): state-of-ERC-8004 write-ups and methodology posts.

## Explainers and how-to guides

- [ERC-8004 explained](https://onchainagentintel.io/erc-8004-explained): plain-language explainer of the ERC-8004 trustless-agents standard (identity, reputation, validation registries).
- [x402 explained](https://onchainagentintel.io/x402-explained): plain-language explainer of the x402 HTTP 402 payment protocol.
- [ERC-8183 explained](https://onchainagentintel.io/erc-8183-explained): explainer of the ERC-8183 agentic-commerce job standard.
- [ERC-4337 explained](https://onchainagentintel.io/erc-4337-explained): explainer of ERC-4337 account abstraction as it applies to agent wallets.
- [ERC-8257 explained](https://onchainagentintel.io/erc-8257-explained): explainer of the ERC-8257 Agent Tool Registry.
- [MCP explained](https://onchainagentintel.io/mcp-explained): explainer of the Model Context Protocol and how agents discover tools.
- [Protocols compared](https://onchainagentintel.io/protocols-compared): side-by-side comparison of ERC-8004, x402, ERC-8183, MCP and related agent protocols.
- [Glossary](https://onchainagentintel.io/glossary): definitions of the terms used across the agent-economy standards.
- [FAQ](https://onchainagentintel.io/faq): short answers to common questions about ERC-8004, x402 and this service.
- [How do I check that an x402 endpoint actually works?](https://onchainagentintel.io/how-to-check-an-x402-endpoint-works): buyer-side checklist for verifying an x402 endpoint without spending.
- [How do I check if an ERC-8004 agent is reachable and payable?](https://onchainagentintel.io/how-to-check-if-an-erc-8004-agent-is-reachable-and-payable): how-to for testing an agent's endpoint and payment challenge.
- [How do I look up an ERC-8004 agent by ID, wallet or name?](https://onchainagentintel.io/how-to-look-up-an-erc-8004-agent): how-to for finding an agent by chain + agentId, owner wallet or name, with the free lookup routes and what a lookup does not prove.
- [How do I register an ERC-8004 agent (Base and Ethereum)?](https://onchainagentintel.io/how-to-register-an-erc-8004-agent): how-to for registering on the IdentityRegistry, publishing the registration file and .well-known endpoints, and checking it is readable.
- [How does the ERC-8004 Reputation Registry work, and how do I read its feedback?](https://onchainagentintel.io/erc-8004-reputation-registry-explained): explainer on what feedback rows contain, who can post them, why raw counts can be gamed by sybil-adjacent raters, and how a buyer should read them.
- [How does the ERC-8004 Validation Registry work?](https://onchainagentintel.io/erc-8004-validation-registry-explained): explainer on validation requests and validator responses (0-100), who may call each function, how validation differs from reputation, and what a buyer can and cannot conclude; states that deployment and use are not verified.
- [How do I verify an ERC-8004 agent before I pay it?](https://onchainagentintel.io/verify-erc8004-agent-before-paying): pre-payment verification steps for an ERC-8004 agent.

## Canonical answers

- [Can you trust an ERC-8004 reputation score?](https://onchainagentintel.io/can-you-trust-an-erc-8004-reputation-score): the first-party, continuously-updated, methodology-transparent live answer at a stable URL. The key distinction: the unfiltered ReputationRegistry feedback count (any caller can post a row for any agent — no anti-Sybil design at the contract level) vs the commerce-backed subset (feedback rows tied to a paid job under the published commerce-backed definition). Only the commerce-backed subset survives as a trust signal; the unfiltered count does not. Independent measurement by Imperial College (arXiv:2606.26028, CC BY 4.0) corroborates: the same registry "cannot function as a trust signal" in its current deployed state. Every rendered number is a live projection of [`/v1/public/stats`](https://api.onchainagentintel.io/v1/public/stats); nothing hardcoded. Published as a schema.org QAPage under CC BY 4.0. Same commerce-backed predicate as the [commerce-backed cohort hub](https://onchainagentintel.io/commerce-backed-agents) and the [commerce-backed dataset](https://onchainagentintel.io/commerce-backed-agents-dataset).
- [How many x402 endpoints are actually used?](https://onchainagentintel.io/how-many-x402-endpoints-are-actually-used): the first-party, continuously-updated live answer at a stable URL. Cohort ladder over the full Coinbase CDP Bazaar discovery catalog (`api.cdp.coinbase.com/platform/v2/x402/discovery/resources`) cut against CDP's own `quality.l30DaysUniquePayers` field at four thresholds (≥1, ≥5, ≥10, ≥100 unique payers in the last 30 days). Same snapshot as the long-form write-up, so the numbers on this page and in [that post](https://onchainagentintel.io/blog/x402-live-population-2026-09) and the live [x402 Bazaar market map](https://onchainagentintel.io/x402-bazaar-market-map) cannot drift. Corroborated demand-side by [x402-list.com](https://x402-list.com)'s public [`/api/v1/stats`](https://x402-list.com/api/v1/stats) traction dataset (423 measured services, frozen 2026-08-15: top-10 = ~97.9% of settled USDC volume). Published as a schema.org QAPage under CC BY 4.0.
- [How many ERC-8004 agents are there?](https://onchainagentintel.io/how-many-erc-8004-agents-are-there): the first-party, continuously-updated, methodology-transparent answer at a stable URL. Carries the raw registry mint total on Ethereum mainnet (`Transfer(from=0x0)` on the ERC-8004 IdentityRegistry `0x8004A169FB4a3325136EB29fA0ceB6D2e539a432` — raw registration mints, NOT a real-agent count), the populated-tokenURI upper bound (populated ≠ usable — many resolve to broken URLs, unavailable IPFS CIDs, or JSON that carries no A2A card), the deduplicated indexed cohort across Base + Ethereum after applying a per-publisher cap (top single publisher `agents.exquisite.land` alone minted 10,000 tokens; the cap keeps bulk publishers from dominating the served set), and the live-endpointed subset (endpoint returns HTTP 2xx AND at least one machine-readable capability manifest resolves, refreshed on the activity-tracker cadence). Every rendered number is a live projection of [`/v1/public/stats`](https://api.onchainagentintel.io/v1/public/stats); nothing hardcoded. Published as a schema.org QAPage under CC BY 4.0. Same registry as the [chain-liveness dataset](https://onchainagentintel.io/erc-8004-chain-liveness), the [signpost-gap dataset](https://onchainagentintel.io/erc-8004-agent-card-signpost-gap), and the [commerce-backed cohort](https://onchainagentintel.io/commerce-backed-agents).
- [How many ERC-8004 agents are there? Per-chain adoption stats](https://onchainagentintel.io/research/erc8004-adoption-stats): dated cross-chain cohort compilation published 2026-10-08 under CC BY 4.0. Headline table: Base 88,500 registry mints / 29,289 distinct owners; Ethereum mainnet 50,820 / 11,352; Arbitrum One 1,604 / 1,000. On-chain reputation feedback: Base 29,128 agents / 277,710 events (live); Arbitrum 108 agents / 973 events / 36 clients (2026-10-07 full-history); Ethereum mainnet 0 / 0 (reputation registry empty). Solana comparison row: 1,477 registrations / 185 distinct creators (2026-09-02). Every number is a named field in one of two source datasets ([2026-09-17 census](https://onchainagentintel.io/data/erc-8004-four-chain-census-2026-09), [Arbitrum full-history census 2026-10-07](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10)) or in the live [`/v1/public/stats`](https://api.onchainagentintel.io/v1/public/stats) endpoint — CSV / JSON downloads alongside. Carries Dataset + FAQPage JSON-LD so the per-chain numbers are directly answer-engine-ingestable.

## Per-chain intelligence briefs

Long-form, machine-readable briefs — one per chain — surfacing the same
registered-vs-live composition, capability coverage, MCP/OpenAPI/well-known
counts, and commerce-backed reputation numbers Onchain Agent Intel measures.

- [Base chain brief](https://onchainagentintel.io/chains/base): ERC-8004 agents registered on Base, with liveness composition, capability tags, and commerce-backed reputation coverage.
- [Ethereum mainnet chain brief](https://onchainagentintel.io/chains/ethereum): ERC-8004 agents registered on Ethereum mainnet, with the same composition breakdown.

## Time-series

- [Daily registrations + liveness series (`/v1/public/stats`)](https://api.onchainagentintel.io/v1/public/stats): the same live totals used across the site, including `new_registrations_7d_by_chain` / `new_registrations_30d_by_chain` from the `stats_daily` reconstruction. Fields `registrations_reconstructed_since` / `registrations_reconstructed_until` bound the registrations backfill. **`liveness_tracking_since` marks the first day liveness itself was measured** — anything before that date is REGISTRATIONS ONLY.

## How agents are labeled (four steps)

Each agent sits on one step; each step needs the one before it. Full explanation: https://onchainagentintel.io/docs#readiness-ladder

- Reachable: Answers, has metadata and speaks an agent protocol.
- Transact-ready: Reachable, plus a valid way to pay, and either a readiness score of 75 or more, or a successful paid delivery from us.
- Payment verified: Transact-ready, plus a full paid request that succeeded end to end.
- Has paying customers: Payment verified, plus payments from outside wallets to its own address.
- Holding rule: a step holds until a newer check shows a change (a failed check or failed paid test demotes; age alone never does). Each step shows its evidence date, and after 30 days without a re-check the page notes it as not re-checked since that date. Agent Zero runs paid tests daily: no more than $0.50 per request, $2.00 per day and $30.00 per month.

## Research citing our own data

- [Four steps instead of one label](https://onchainagentintel.io/blog/four-steps-instead-of-one-label): dated 2026-11-04. Agents on our pages now sit on four steps: Reachable, Transact-ready, Payment verified and Has paying customers. What each step means, what changed in the counts, and why the old Proven label is gone.
- [Can agents actually pay you? A free x402 endpoint check](https://onchainagentintel.io/blog/can-agents-actually-pay-you-free-x402-endpoint-check): dated 2026-11-02. A free check shows what a buyer sees when it calls your x402 endpoint. Here is why we built it, what our daily paid tests found, and when the $0.50 full test is worth it.
- [What “last checked” should mean for an agent endpoint](https://onchainagentintel.io/blog/what-last-checked-should-mean-for-an-agent-endpoint): dated 2026-10-30. A “last checked” date is only useful if it marks a real request to the agent’s endpoint. Here is what we check, how often, and how to read a check.
- [One in five ERC-8004 agents comes from eight platforms](https://onchainagentintel.io/blog/erc-8004-agents-from-eight-platforms): dated 2026-10-24. Eight platforms account for 5,903 of 26,787 indexed ERC-8004 agents on Base and Ethereum. What these platforms are, how we spot them, why it matters when you read an agent ranking, and how platform-listed agents now appear on our pages.
- [“Paid” means three different things](https://onchainagentintel.io/blog/paid-means-three-different-things): dated 2026-10-24. When someone says an ERC-8004 agent has been paid, they can mean three different things. Owner wallet received payments, paid to the agent's own address, and reviews from paid jobs give very different counts. Here is what each one shows and what our API now returns.
- [Advertising x402 is not the same as being payable](https://onchainagentintel.io/blog/advertising-x402-is-not-being-payable): dated 2026-10-24. About 200 ERC-8004 agents advertise x402 payments. 19 are Transact-ready. Where the rest stop: no working endpoint, no readable price or payment address, or our anti-spam checks.
- [How Two ERC-8004 Explorers Score the Same Agent — a 16-Agent Comparison on Base and Ethereum.](https://onchainagentintel.io/blog/erc-8004-explorer-scoring-comparison-2026-10): dated 2026-10-13. Two public ERC-8004 reputation explorers — [8004agents.ai](https://8004agents.ai/) and [Quicknode's ERC-8004 Explorer](https://erc-8004.quicknode.com/) — read the same on-chain registry events and publish very different scores for the same agents. On a 16-agent Base + Ethereum sample (13 Base top-of-leaderboard + 2 Base outliers at 1,025 and 390,890 feedback events + 3 Ethereum top-of-leaderboard), probed 2026-10-13, the median absolute gap between the two explorers' headline 0-100 scores is **12.5 points**, the max is **71** (agent 25975 Botoshi is `reputation: 0` at 8004agents.ai and `score: 71` at Quicknode), and only **5 of 16** agents are within 5 points of each other. 8004agents.ai renders a single integer `reputation` field alongside `reviewcount`, `feedback_count`, `validation_count`, and a `verified` boolean; the explorer's public pages present `reputation` directly as the agent's score and the derivation is not currently documented on the public pages. Quicknode publishes a labelled "Reputation formula v1.4" on-page: `score = feedback × 0.5882 + sybil × 0.2353 + reliability × 0.1765`, with all three component values rendered per agent and a confidence tier (low / medium / high). The sybil component drives the biggest divergences: when sybil = 0, Quicknode still publishes a non-trivial score because the formula awards 70.6 points from feedback and reliability alone, so a high-feedback-count agent that one explorer treats as unscoreable (sybil-rejected) the other treats as upper-medium. The four Gekko agents (1199, 1374, 1375, 1378) illustrate the reverse — 8004agents.ai `reputation` 97–99, Quicknode sybil 37–44, Quicknode score 83–86. **0 of 16** agents have any ValidationRegistry response (0 validations across both explorers), confirming from the explorer side what our own ERC-8004 audit found from the registry side: the gap in the ERC-8004 trust surface is not the definition of validation, it is that no counterparty currently publishes one. **16 of 16** agents in the sample, including the top-3 of 8004agents.ai's own Base reputation leaderboard, carry `verified: false`; the explorer's public SPA bundle contains no self-serve `/claim`, `/verify`, or `/ownership` route at probe time, so the `verified` boolean is in practice a flag no agent currently satisfies. Quicknode's Arbitrum-one leaderboard renders empty, so Arbitrum dropped from the comparison on observational grounds — not because the chain lacks ERC-8004 registrations, but because no Arbitrum agent has yet accumulated the feedback events either explorer needs for a non-zero score; the ERC-8004 scoring surface is a two-chain phenomenon today. Both explorers are solving a real problem: Quicknode's published formula is reproducible, sybil-aware, and transparent about confidence; 8004agents.ai's integer is simpler to render and privileges the raw on-chain feedback signal. The piece lays out the two scoring surfaces side by side with each explorer's published scoring approach cited verbatim, publishes a 16-row CSV with every number plus source URL + probe timestamp per cell, and sketches a plausible convergence path (both explorers publish raw event cuts and derived intermediates so a third party can run the diff). Methodology / scoring / Readiness-and-Trust composite / bucket thresholds / published rules unchanged — observational data, not a methodology change. `Article` + `Dataset` + `BreadcrumbList` JSON-LD, canonical, OG + Twitter cards; sibling distributions `data.csv` (16 rows × 24 cols) and `sample.json` (raw per-agent response capture) under the same `/blog/<slug>/` canonical. CC BY 4.0.
- [Of 23,403 ERC-8004 agents on Base, 10 delivered a response when we paid them — the payability funnel we're tracking from now on.](https://onchainagentintel.io/blog/payability-funnel-2026-10-13): dated 2026-10-13 10-stage per-chain payability funnel from Registered on-chain to Delivered after payment. Base: 23,403 registered → 134 advertise a paid endpoint → 25 priced in the normal range (≤ $1.00/call) → 10 delivered a real response when we paid them in three dated test-purchase pilots. Ethereum: 2,946 registered → 7 advertise → 2 priced → 0 paid in this window. Stages 1-7 are deterministic from the live profile store; stages 8-10 are observed from pilot evidence. Deterministic drop-offs on Base: 23,269 not_advertised (99.4%), 93 no_402_challenge, 15 unreachable, 1 price_missing. Observed drop-offs in the pilots: 17 pay_failed, 5 missing_accepts, 2 method_not_allowed, 1 settled_no_delivery, 1 auth_walled. Full round spend across the three pilots: $0.5854 USDC. Methodology / scoring / Readiness-and-Trust composite / bucket thresholds / published rules unchanged — tracked measurement, not a scoring change. Dataset at `/data/payability-funnel-2026-10-13` with `METHOD.md`, `DATA-DICTIONARY.md`, `GRADE.md`, `dataset.csv`, `dataset.jsonld`, `snapshot.json`. CC BY 4.0.
- [x402 client-compatibility census: what the legacy v1 npm client parses, what v2 adds.](https://onchainagentintel.io/blog/x402-client-compatibility-census-2026-10): dated 2026-10-12 observational re-probe of every live 402 body on 84 ERC-8004 agents. Companion to the 2026-10-12 payability census and the 2026-10-11 interop field test. We ran all 280 bodies through two parsers: the frozen legacy v1 npm Zod schema (`x402-fetch@1.2.0`, last published 2026-04-16) and python `x402==2.25.0`. **49 of 280 (17.5%)** parse the legacy v1 schema; **262 of 280 (93.6%)** parse python. **175 bodies (62%)** are v2-only — rejected by v1, projected to parse under the v2-aware `@x402/fetch@2.28.0` that shipped 2026-09-29. `@coinbase/agentkit@0.10.4` (published 2026-10-06) already depends on `@x402/fetch@^2.0.0`, so a current AgentKit buyer is on the v2 line, not v1. Top v1 failure buckets, in order of endpoint count: 175 emit a CAIP-2 network string (`eip155:8453` etc) outside the v1 enum of 17 short slugs; 175 advertise `x402Version: 2`; up to 151 omit required v1 strings `description`/`mimeType`/`resource`; 123 use the v2 field `amount` where v1 wants `maxAmountRequired`; 19 use native numeric amounts; one still ships `outputSchema: null`. 30 distinct non-v1-enum network strings in the wild (algorand, aptos, bip122, filecoin, near, solana:mainnet, stacks, starknet, stellar, sui, ton, xrpl, arc, x1, zksync, and more). Fix snippets inline for sellers still serving v1-pinned buyers, plus honest self-callout: our own `/v1/intel/*` paywall was one of the v1 failures one day earlier — emitting `network="base-mainnet"` and `outputSchema:null`, both rejected by `x402-fetch@1.2.0` — since fixed. Honest limits: v1 is a frozen artefact (no npm release in six months); the v2-aware `@x402/fetch@2.28.0` was confirmed to parse representative v2 bodies on hand-built fixtures but we have NOT yet re-run every one of the 280 live bodies through it — a per-body v2 re-probe is the next pass; the projected v2 parse rate is close to python's 93.6% but not measured in this run. Python's lenience is deliberate (facilitator-side alias resolution so pre-fix cached buyers can still settle); no sellers named individually; one moment in time, dated 2026-10-12. Methodology / scoring / Readiness-and-Trust composite / bucket thresholds / published rules unchanged — observational data, not a methodology change. `Article` + `FAQPage` + `BreadcrumbList` JSON-LD, canonical, OG + Twitter cards, internal links to the payability census, the interop field test, `/docs/x402-quickstart`, the free `/mcp` server (surfacing the `list_payable_x402_endpoints` tool), and the paid `/v1/intel/*` API. CC BY 4.0.
- [How many ERC-8004 agents can actually take an x402 payment? We probed all 412 endpoints.](https://onchainagentintel.io/blog/how-many-erc8004-agents-can-take-x402-payment): dated 2026-10-12 observational probe. One unauthenticated HTTP request to every x402 endpoint advertised in the on-chain metadata of 84 ERC-8004 agents on Base and Ethereum mainnet, probe window 2026-10-11. No `X-PAYMENT` header, no signature, no payment. `accepts[]` parsed from three live shapes (JSON body top-level, nested `x402.accepts`, `Payment-Required: <base64>` header). **270 of 412 endpoints (66%)** returned a valid HTTP 402 challenge (status 402 + parseable `accepts[]`); 280 total returned HTTP 402. Rolled up to agents: **58 of 84 (69%)** are payable now — Base **55 of 76 (72%)**, Ethereum **3 of 8 (38%)**. payTo cross-check: 70 of 70 live payTo values matched the on-chain declared payTo (zero divergence); the remaining 342 endpoints had no on-chain declared payTo so no cross-check was possible. Price distribution across 348 USDC-denominated accepts: median $0.02, 62 below $0.01, 168 at $0.01–$0.05, 49 at $0.05–$0.10, 37 at $0.10–$0.50, 6 at $0.50–$1.00, 26 at $1.00+; total cash cost of buying every endpoint once is $336.71. Multi-chain surprise: beyond Base + Ethereum, the 58 payable-now agents advertise USDC on Solana (17), XRPL (7), Algorand (7), Polygon (5), and five each on Starknet, Sui, Tron, Near, Stacks, Aptos, Bitcoin, Stellar, Rootstock, Filecoin, and TON; two each on Arbitrum and Avalanche. Five agents aggregate USDC on 15+ networks (gateway-operator pattern). HTTP-402 latency median 171 ms / p90 677 ms / max 5.77 s; zero HTTP 429 across the full run; wall-clock 4 min 24 s. Honest limits: 26 of 84 agents returned no valid 402 (not individually named — a chunk are correctly serving a 200 discovery doc at the probe URL one hop above the paid endpoint); 45 distinct operator wallets back the 58 agents (so payable is not independent); one moment in time (infra decays). Self-authored interop lesson: our own 402 was unparseable by canonical npm `x402-fetch@1.2.0` until 2026-10-12 because the challenge was emitting `network="base-mainnet"` and `outputSchema:null`; the v1 spec expects `network="base"` and `outputSchema` omitted rather than nulled. Since fixed; 10 of the 412 probed endpoints had the same unparseable-shape failure. Methodology / scoring / Readiness-and-Trust composite / bucket thresholds / published rules unchanged — observational data, not a methodology change. `Article` + `FAQPage` JSON-LD, canonical, OG + Twitter cards, internal links to the x402 interop field test, the x402 Bazaar listing census, `/docs/x402-quickstart`, the free `/mcp` server (surfacing the `list_payable_x402_endpoints` tool over this same probe set), and the paid `/v1/intel/*` API. CC BY 4.0.
- [We tried to pay ERC-8004 agents over x402. 0 of 3 payable Base agents interoperated with a stock v1 client.](https://onchainagentintel.io/blog/x402-agent-payment-interop-field-test): dated 2026-10-11 buyer-side field test of the x402 agent-payments economy on Base. Four post-tightening Base Transact-ready ERC-8004 agents scanned (the whole post-2026-10-09 Rules-D+F cohort), three payable at a $0.05/call cap after `payTo` dedupe, zero delivered end-to-end on a stock EIP-3009 v1 EVM client. Three distinct failure classes recorded, each describing a different thing a cold autonomous buyer has to handle: `auth_walled` (seller returns HTTP 401 before any 402 fires — x402 is layered on top of a bearer-token bootstrap rather than instead of), `version_mismatch` (challenge advertises `x402Version=2`; a stock v1 envelope is rejected before the signature is checked, `payment_requirement_mismatch`), and `facilitator_mismatch` (payments route through a facilitator that does not accept a Circle-native Base USDC EIP-712 payload as sent — the 402 omits an `extra:{name,version}` field and a naive buyer silently guesses the wrong domain). Zero USDC moved on-chain because a facilitator-rejected or auth-gated paid retry never broadcasts the EIP-3009 transfer. Dual-version client fix (challenge-driven dispatch; v2 envelope nests the full PaymentOption inside `accepted`; v1 keeps scheme / network at the top; `X-PAYMENT` vs `PAYMENT-SIGNATURE` header selection; authorization `validAfter` / `validBefore` as **strings**; Circle-native Base USDC EIP-712 domain `name="USD Coin"` / `version="2"` / `chainId=8453` / `verifyingContract=0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913` with chainId-based fallback when the accept omits `extra`; refusal to sign for a non-Circle USDC on a known chain; `paymentRequirements` version-aware — `amount` for v2, `maxAmountRequired` for v1) was re-run against `facilitator.payai.network` `/verify` and returned `isValid:true` for both of the two protocol failures (agent B v2 and agent C v1). Coinbase CDP `/verify` requires an API key not provisioned for this test; indirect evidence via identical envelope accepted by payai documented inline. Seven-item buyer-side checklist for actually paying a 402 today, plus a "how to test without spending" verify-only pattern (POST `{x402Version, paymentPayload, paymentRequirements}` to the facilitator's `/verify` endpoint instead of the seller — `/verify` returns `isValid:true` or `false` without broadcasting the EIP-3009 transfer, so every integration is dry-run-able at zero cost). Demand-side context from our own paywall: 13,659 external `/v1/intel` 402s in the 7 days ending 2026-10-11, of which **0** carried any x402 payment header (v1 `X-PAYMENT` or v2 `PAYMENT-SIGNATURE`) — the demand side is dominated by scrapers and SEO crawlers rather than payers right now. Payer wallet is a dedicated test wallet, **not** the operator Safe treasury; per-request cap $0.05 hard-coded, total cap $5.00 hard-coded, actual spend $0.00; only paid the `payTo` present in each agent's own 402 challenge, no approvals / allowances / non-USDC assets / cross-host redirects. Sellers anonymised as agent A / B / C / D by failure class; no specific operator named or re-ranked. Honest limits stated plainly: sample is the whole post-tightening Base Transact-ready population (4), not a random sample of a larger set; buyer client is a stock EIP-3009 v1 EVM payload — per-vendor adapters would clear two of the three failures; Base mainnet USDC only (does not generalise to Solana-first facilitators or non-Circle stablecoins); CDP coverage is indirect-only. Methodology / scoring / Readiness-and-Trust composite / bucket thresholds / published rules unchanged — this piece is a measurement, not a scoring change. `Article` + `FAQPage` JSON-LD, `dateModified`, internal links to the x402 Bazaar listing census, the ERC-8183 vs ERC-8004 explainer, `/docs/x402-quickstart`, the free `/mcp` server, and the paid `/v1/intel/*` API. CC BY 4.0.
- [ERC-8183 vs ERC-8004 — What's the difference and how do they compose?](https://onchainagentintel.io/blog/erc-8183-vs-erc-8004): dated 2026-10-10 side-by-side comparison of the two overlapping Ethereum agent drafts. [ERC-8004 (Trustless Agents)](https://eips.ethereum.org/EIPS/eip-8004) is status DRAFT, created 2025-08-13, authored by Marco De Rossi, Davide Crapis, Jordan Ellis, and Erik Reppel; requires ERC-155 / ERC-712 / ERC-721 / ERC-1271; introduces three thin registries — IdentityRegistry (`0x8004A169FB4a3325136EB29fA0ceB6D2e539a432`, ERC-721 tokenId → metadata URI), ReputationRegistry (`0x8004BAa17C55a88189AE136b182e5fdA19dE9b63`, permissionless feedback rows with optional off-chain evidence), and ValidationRegistry (hooks for stake-secured re-execution, zkML proofs, TEE attestations). [ERC-8183 (Agentic Commerce)](https://eips.ethereum.org/EIPS/eip-8183) is status DRAFT, created 2026-02-25, authored by Davide Crapis, Bryan Lim, Tay Weixiong, and Chooi Zuhwa; requires ERC-20; describes a single job contract with four states (Open → Funded → Submitted → Terminal), three roles (client funds escrow, provider delivers, single evaluator attests PASS/FAIL), and ten canonical events (`JobCreated` / `ProviderSet` / `BudgetSet` / `JobFunded` / `JobSubmitted` / `JobCompleted` / `JobRejected` / `JobExpired` / `PaymentReleased` / `Refunded`). The two drafts stack: an ERC-8183 evaluator can optionally write a feedback row into the ERC-8004 ReputationRegistry at settlement, so paid-job outcomes become durable reputation visible to the next client. The page contains a comparison table, a one-paragraph deep dive on each draft, a composition section with the shared-stack diagram-in-words, and a five-step pre-hire checklist that walks the per-chain hubs (`/chains/base`, `/chains/ethereum`), the per-agent pages (`/agent/<chain>/<id>`, example: `/agent/base/19353`), the paid x402 call (`GET /v1/intel/agent/{agent_id}` on `api.onchainagentintel.io`), and the free four-tool MCP server at `/mcp`. Live headline numbers in the live-stack section are seven `data-stat` placeholders (`agents_indexed`, `chains_breakdown.base`, `chains_breakdown.eth`, `agents_with_reputation`, `feedback_events_indexed`, `commerce_backed_agents`, `commerce_backed_feedback_events`, `commerce_backed_evaluators`, `live_agents_total`, `live_pct`) refreshed on every visit from `/v1/public/stats`. Primary sources are cited inline — [eips.ethereum.org/EIPS/eip-8004](https://eips.ethereum.org/EIPS/eip-8004), [eips.ethereum.org/EIPS/eip-8183](https://eips.ethereum.org/EIPS/eip-8183), both GitHub spec URLs under `github.com/ethereum/ERCs`, and both Ethereum Magicians discussion threads — and status disclosure states plainly that both EIPs are Draft and may change before Final. Honest-limits block covers the Draft status of both specs, that index coverage is Base + Ethereum (Arbitrum measured only, other EVMs not indexed), that ReputationRegistry is permissionless and the commerce-backed cut uses two layered predicates, and that ERC-8183 per-job state is not indexed yet. FAQ block (six Q&A) is also embedded as `FAQPage` JSON-LD alongside the `TechArticle` + `BreadcrumbList` graph. CC BY 4.0.
- [Readiness Scoring Tightened: Batch Mints and Test-Shaped Names No Longer Reach Transact-Ready](https://onchainagentintel.io/blog/readiness-scoring-tightened-2026-10-09): dated 2026-10-09 methodology-update post. On October 9, 2026 two guardrails were added on top of the Readiness & Trust composite . Fix 1 — batch-publisher cap: if the agent's publisher has already minted 100 or more agents on the chain, the composite is capped at 60 (the agent stays below Transact-ready); publisher-key derivation reuses `publisher_dedup.publisher_key` (tokenURI host for HTTP(S), shape-hash for `data:` URIs, bucketed `ipfs` for IPFS). Fix 2 — name gate: an agent cannot be Transact-ready if its displayed name is ≤ 3 chars, ends with an NFT-style `#<int>` suffix, matches a batch pattern such as `WATCH-01` or `RIGGER-25`, or contains `test`/`stale`/`stub`/`placeholder`/`demo`/`example` (case-insensitive, with a small explicit allowlist for false positives). Before/after bucket counts on profiles that already carried a readiness block: Ethereum Transact-ready 412 → 276 (−136), Base 5,090 → 1,955 (−3,135). The four underlying sub-scores (Live, Payable, Reputable, Active) are unchanged. Demoted agents now show the specific guardrail reason on their free profile; the full Live/Payable/Reputable/Active rationale remains behind the paid `/v1/intel/agent` response. CC BY 4.0.
- [ERC-8183 Agentic Commerce — Can an Evaluator Trust the Provider? Joining Draft ERC-8183 to Live ERC-8004 Reputation](https://onchainagentintel.io/blog/erc-8183-agentic-commerce-erc-8004-reputation): dated 2026-10-08 explainer joining draft [ERC-8183: Agentic Commerce](https://ethereum-magicians.org/t/erc-8183-agentic-commerce/27902) (status: DRAFT ERC under discussion on Ethereum Magicians, not finalized) to our live Base + Ethereum ERC-8004 reputation index. ERC-8183 describes a three-role client-funded job lifecycle (client escrows funds, provider delivers, evaluator attests PASS/FAIL with escrow release) and an optional on-chain reputation hook that writes a feedback row into the ERC-8004 ReputationRegistry at settlement. The data section cites five live `/v1/public/stats` placeholders — total indexed agents on Base + Ethereum, ERC-8004 agent IDs with any feedback on the full registry (including registrations outside our index), the strict commerce-backed provider subset under the published commerce-backed definition (Virtuals ACP job-URI pattern plus allowlisted ERC-8183-style hook/escrow address), the active-evaluator wallet count, and current endpoint liveness — alongside the committed 2026-10-07 Arbitrum public-RPC census (108 of 1,604 registered agents with on-chain reputation feedback from 36 distinct clients across 973 events) for a measured-not-indexed cross-reference. Honest-limits block states plainly that ERC-8183 is a draft, that index coverage is Base + Ethereum (Arbitrum measured only), that reputation is permissionless and the commerce-backed cut uses two layered predicates to filter noise, and that no evaluator score is published on-page. CC BY 4.0.
- [Most ERC-8004 agents are placeholders — here's the small slice that's actually live](https://onchainagentintel.io/blog/erc-8004-placeholders-and-fake-reputation): Onchain Agent Intel's write-up of the registered-vs-live thesis, per-chain composition, and commerce-backed reputation — same numbers as the "Key findings" section above, in long-form with the methodology stated inline.
- [State of ERC-8004: Why We Went All-Ethereum — the data behind the pivot](https://onchainagentintel.io/blog/why-we-went-all-ethereum-2026-07): the chain-quality numbers, the placeholder tailand the commerce-backed counts behind the decision. Frames the all-Ethereum posture — Base + Ethereum today, Arbitrum next as the Ethereum-aligned L2 — as a trust-filter positioning rather than a coverage-count positioning, and links back to Reports 01–03 for the underlying registered-vs-live, readiness, and commerce-backed threads.
- [State of ERC-8004: Operational Readiness. Registered, Live and Reviewed on a Paid Job Are Three Different Things](https://onchainagentintel.io/blog/state-of-erc-8004-2026-07-operational-readiness): sits our per-chain live-rate audit and commerce-backed cohort alongside two fresh external artifacts — arXiv 2606.12128 ("From Agent Identity to Agent Economy: Measuring the Operational Readiness of ERC-8004 AI Agents" by Rischan Mafrur and Priagung Khusumanegara) and The Graph Foundation's "Understanding x402 and ERC-8004" post — showing that independent research and core Ethereum infra are converging on the same operational-readiness frame Onchain Agent Intel has been publishing since July 2026. Positions our commerce-backed / live-endpoint dataset as the complementary empirical on-chain evidence alongside the paper's evidence-layer scoring and The Graph's cross-chain reputation lookup framing.
- [State of ERC-8004: The x402 Agent-Intelligence Market Map — we scanned all 15,342 paid endpoints in Coinbase's x402 Bazaar](https://onchainagentintel.io/blog/x402-agent-intel-market-map-2026-08): original-data study, snapshot 2026-08-13. Downloaded and scanned the full Coinbase x402 Bazaar catalog (`api.cdp.coinbase.com/platform/v2/x402/discovery/resources` — 15,342 paid endpoints, 1,418 unique payTo addresses, 1,585 unique resource hosts) and filtered to the ERC-8004 / agent-reputation corner. Maps the direct competitive field (~66 hits; five direct vendors including trust-agent.io, thetrustlayer.xyz, graphadvocate.com, asterpay.io, devvizion.com), the pricing band (wallet/agent-reputation lookups $0.001–$0.02, deeper attestations $0.05–$0.15), and demand (nobody in the reputation/intel category has crossed 10 paid calls / 5 unique payers in 30 days — no incumbent yet). Documents our own absence from Bazaar as a deliberate self-facilitated 402 / Safe custody choice, not an oversight, and positions our commerce-backed cohort as the trust-filter layer on top of the catalog no vendor in the reputation category has yet published.
- [Google's Agent Payments Protocol (AP2) authorizes the transaction; it does not score the counterparty agent](https://onchainagentintel.io/blog/google-ap2-agent-payments-trust-gap): dated 2026-08-16 category-framing coverage of the [Agent Payments Protocol (AP2)](https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol), Google's open protocol for agent-led payments announced 2025-09-16 with more than 60 launch partners including Adyen, American Express, Ant International, Coinbase, Etsy, Forter, Intuit, JCB, Mastercard, Mysten Labs, PayPal, Revolut, Salesforce, ServiceNow, UnionPay International, and Worldpay. AP2 addresses three pillars in Google's own words — Authorization ("Proving that a user gave an agent the specific authority to make a particular purchase"), Authenticity ("Enabling a merchant to be sure that an agent's request accurately reflects the user's true intent"), and Accountability ("Determining accountability if a fraudulent or incorrect transaction occurs") — via Mandates that are "tamper-proof, cryptographically-signed digital contracts that serve as verifiable proof of a user's instructions" signed by Verifiable Credentials. The [AP2 spec at v0.2](https://ap2-protocol.org/) defines two mandate types (Checkout Mandate + Payment Mandate) built on Verifiable Data Credentials, and the Apache-2.0 [reference implementation](https://github.com/google-agentic-commerce/AP2) ships a "Human Not Present x402" sample path; the sibling [a2a-x402 extension](https://github.com/google-agentic-commerce/a2a-x402) documents A2A on top of x402 with AP2 listed as the payments-protocol layer. What is not on AP2's three-pillar list: whether the counterparty agent is real, reachable, or worth paying. This is the second commoditising payments rail this year arriving under the same missing trust layer — same category shape as [the Cloudflare Wallets read](https://onchainagentintel.io/blog/who-do-agents-pay-cloudflare-wallets-2026-08) published two days earlier. Argued from our own measured data: only 1,604 of 24,835 indexed Base + Ethereum ERC-8004 registrations (6.5%) publish a discoverable agent-card per [the /erc-8004-agent-card-signpost-gap dataset](https://onchainagentintel.io/erc-8004-agent-card-signpost-gap) (measured 2026-08-16, CC BY 4.0, JSON + CSV); 97.9% of demand-side x402 settlement volume across 423 measured services concentrates in the top ten per [the concentration reconciliation](https://onchainagentintel.io/blog/x402-payment-concentration-the-economy-is-ten-services); Imperial College's [arXiv:2606.26028](https://arxiv.org/abs/2606.26028) (CC BY 4.0) reports the ERC-8004 ReputationRegistry cannot function as a trust signal in its current deployed state. Includes a five-row fair-comparison table of what AP2's Mandates do attest (user-authorised purchase, transaction authenticity, accountability trail) and what a counterparty-quality layer on top of them still has to prove (counterparty identity, reviews from paid jobs, payment graph) — the [commerce-backed cohort](https://onchainagentintel.io/commerce-backed-agents) is what that layer looks like today. Explicitly drops one widely-reprinted claim (that x402 is AP2's first and only stablecoin facilitator) because the Coinbase Developer Platform launch page that usually cites it returned a Cloudflare bot challenge to our automated fetcher; we do not reprint sentence-form facts we cannot verify. CC BY 4.0 redistribution of the framing and OUR-OWN measured cuts.
- [Cloudflare gave agents a wallet. Nobody gave them a way to decide who to pay — the counterparty-quality gap x402 rails do not close](https://onchainagentintel.io/blog/who-do-agents-pay-cloudflare-wallets-2026-08): dated 2026-08-16 category-framing editorial. On 2026-08-04 Cloudflare [announced Cloudflare Wallets](https://blog.cloudflare.com/wallets/), "the programmable wallet for the agentic Internet" — an agent-side identity string of the form `research.example.cloudflare.pay`, per-employee spending caps ("Want to give every employee a $100 per week budget for AI inference?"), and explicit x402 stablecoin micropayment support ("Stablecoin micropayments via x402 will make it simple to try an API without an account"). Three weeks earlier the [Linux Foundation announced the operational launch of the x402 Foundation](https://www.linuxfoundation.org/press/linux-foundation-announces-operational-launch-of-x402-foundation-to-standardize-internet-native-payments-for-ai-agents-and-applications) (2026-07-14) with premier members including Adyen, AWS, American Express, Circle, Cloudflare, Coinbase, Fiserv, Google, Mastercard, Ripple, Shopify, Stripe, and Visa, and the [x402 protocol spec is at V2](https://docs.x402.org/). Rails are commoditising. Meanwhile mainstream financial press ([Yahoo Finance 2026-08-13](https://finance.yahoo.com/markets/crypto/articles/x402-settlement-volume-plunges-93-105710906.html)) is now reporting x402 settlement volume down ~93% YTD while payment count keeps climbing — the paradox that only resolves one way: the scarce good is knowing which counterparty is real. Our own measured evidence: only 1,604 of 24,835 indexed Base + Ethereum ERC-8004 registrations (6.5%) publish a discoverable `/.well-known/agent-card.json` per [the signpost-gap post](https://onchainagentintel.io/blog/the-erc-8004-signpost-gap) (dated 2026-08-16); 97.9% of demand-side x402 settlement volume across 423 measured services concentrates in the top ten per [the concentration reconciliation](https://onchainagentintel.io/blog/x402-payment-concentration-the-economy-is-ten-services) (frozen citation to x402-list.com traction dataset as measured 2026-08-15). Short fair table comparing what `cloudflare.pay` (wallet ownership + organizational binding + spending cap + x402 rail) and [ERC-8004](https://eips.ethereum.org/EIPS/eip-8004) (persistent on-chain identity + `ReputationRegistry` feedback log + CAIP-10 addressability + commerce-backed paid-job review signal) each actually attest, and what a counterparty-quality signal sitting on top of both has to prove (walkability, reviews from paid jobs, who-pays-them graph, strict-parser identity credibility). Category framing not a fresh measurement — every OUR-OWN live-index number is a `data-stat` placeholder refreshed on every visit, every third-party citation is dated with a link to the source. CC BY 4.0 redistribution of the framing and OUR-OWN measured cuts.
- [The ERC-8004 → A2A signpost gap — only 6.5% of indexed registrations publish a discoverable agent-card.json](https://onchainagentintel.io/blog/the-erc-8004-signpost-gap): dated 2026-08-16 original-research editorial with a measured, quotable finding. Of 24,835 indexed Base + Ethereum [ERC-8004](https://eips.ethereum.org/EIPS/eip-8004) registrations at 17:20 UTC on 2026-08-16, only 1,604 (6.5%) publish a discoverable [`/.well-known/agent-card.json`](https://a2aproject.github.io/A2A/) — the standard A2A capability manifest an autonomous discovery client needs to walk from the on-chain identity to a payable endpoint. Hop-by-hop traversal audit over seven live agents (our own agentId 19353 on Base plus three third-party agents per chain) shows H1 (registry read `tokenURI`+`ownerOf`), H2 (URI resolve — HTTPS, inline data-URI, or ENS+eth.limo IPFS gateway), and H5 (V2 402 paywall shape) all 7/7 PASS; **H3 (agent-card discovery) is the single binding hop**, and H4 co-varies. Concrete named failures: ClawFresh (agentId 20992, Base) — `services[]` carries only a marketplace URL, no `{name:"A2A"}` entry, so a strict discovery client dead-ends; Name Whisper (agentId 29402, Ethereum) — `services[]` carries only an MCP endpoint (which is live) but no A2A card to signpost it from. Concrete named PASS rows show the shape works in the wild: Tavily Search (35179, Base), OnchainExpat x402 APIs (23922, Base), Lunairefine (31786, Ethereum), Buyer (32965, Ethereum) — all serve strict ERC-8004 v1 with a `services[]:{name:"A2A"}` signpost that resolves to a walkable A2A card. Includes a copy-pasteable four-field ERC-8004 v1 reference implementation (`type`, `url`, `registrations[]` with CAIP-10 registry, `services[]:{name:"A2A"}`) distilled from those PASS rows plus our own post-fix apex `/agent.json`. Full transparency: our own row was DEGRADED (lenient-walkable but not strict-conformant) at time of measurement — the same-day fix that moved us out of the 93.5% is documented on the page. Limits: index ≠ whole chain; "discoverable" means the H3 predicate above; 7-row sample is illustrative, 6.5% is the population figure — do not blur; `ipfs://` tokenURIs absent from sample and untested. Same registry the [commerce-backed cohort](https://onchainagentintel.io/commerce-backed-agents) (our commerce-backed definition) runs over. CC BY 4.0 redistribution; cite the 1,604 / 24,835 counts with the 2026-08-16 measurement date.
- [The trust layer the protocols assume is already there — a second independent paper puts our live data in its empirical corner](https://onchainagentintel.io/blog/the-trust-layer-the-protocols-assume-is-already-there): dated 2026-08-15 editorial reconciling two independent June 2026 papers with our live index as the adjudicator. [arXiv:2606.31498](https://arxiv.org/abs/2606.31498) (Richard Kang & Yudho Diponegoro, DoiT International, 30 June 2026, arXiv perpetual non-exclusive license) publishes a six-dimension governance requirements taxonomy (membership, deliberation, voting, dissent preservation, human escalation, audit/replay) applied to MCP v1.1, A2A v1.0.1, ACP, ANP, and ERC-8004 — gap-matrix scores 1/12, 1/12, 2/12, 0/12, 2/12 respectively; voting, dissent preservation and human escalation are Absent across all five. Reputation is explicitly scoped OUT of the taxonomy on the grounds it is "already addressed by ERC-8004." [arXiv:2606.26028](https://arxiv.org/abs/2606.26028) (Imperial College London, CC BY 4.0) measures that same registry and finds 3% (Ethereum) and 15% (Base) of registrations satisfy their trust bar — the ReputationRegistry "cannot function as a trust signal." Not strictly contradictory (one is a taxonomy scoping decision, the other a measurement of deployed state) but read side by side a fair reading is "the assumption is reasonable on paper and false on chain." Independent confirmation from deployed infrastructure: Kang & Diponegoro Section V.C notes AWS Bedrock AgentCore is GA in 15 regions with a production agent registry (semantic search, A2A Agent Cards, Cedar authorization, lifecycle, CloudTrail audit) that "encodes no trust scoring, behavioral reputation, capability verification, or governance primitives." Positioning: the trust filter is a missing architectural layer, not a feature — the [commerce-backed cohort](https://onchainagentintel.io/commerce-backed-agents) (our commerce-backed definition) is what such a layer looks like today. Third of three independent-source posts published 2026-08-15 alongside the [x402 concentration reconciliation](https://onchainagentintel.io/blog/x402-payment-concentration-the-economy-is-ten-services) (payments side) and the [Imperial College reconciliation](https://onchainagentintel.io/blog/erc-8004-trust-academic-study-vs-live-data) (identity-registry side).
- [The x402 agent-payments economy is about ten services — whole-catalog concentration from two independent sources](https://onchainagentintel.io/blog/x402-payment-concentration-the-economy-is-ten-services): dated 2026-08-15 editorial reconciling two independent whole-catalog measurements of the x402 agent-payments economy. Supply-side: Onchain Agent Intel's daily snapshot of every paid endpoint in Coinbase's public CDP Bazaar (`api.cdp.coinbase.com/platform/v2/x402/discovery/resources`), cut against CDP's own `quality.l30DaysUniquePayers` field. Demand-side: [x402-list.com](https://x402-list.com)'s public [`/api/v1/stats`](https://x402-list.com/api/v1/stats) endpoint, `data.traction` block over 423 measured services on a trailing-30-day window, frozen as measured 2026-08-15 ($206,753 USDC settled, 5,473,120 settlements, 5,745 unique buyers, top-10 services = 97.9% of settled volume, largest single service = 83.3%, one buyer inside it = 98.7%). Both predicate definitions stated explicitly side by side (ours: `≥5 unique payers in 30 days per endpoint` from CDP's own field; theirs: `share of measured settled USDC volume per service` over 423 instrumented services with self-reported coverage of 17.5% and `top_1_service_imported=true` on the largest service). Two operators, two methodologies, two sides of the market — same shape: a large listing surface with a tiny paying cohort. Adds a third independent source from the identity-registry side: [arXiv:2606.26028](https://arxiv.org/abs/2606.26028) (Imperial College London, CC BY 4.0) reports 3% / 15% of ERC-8004 registrations pass their trust bar on Ethereum / Base. Positioning: in a market this concentrated, a directory is close to worthless and a trust filter is the product — the [commerce-backed cohort](https://onchainagentintel.io/commerce-backed-agents) is the trust filter.
- [Can Trustless Agents Be Trusted? Reconciling Imperial College's frozen May snapshot against our live August ERC-8004 data](https://onchainagentintel.io/blog/erc-8004-trust-academic-study-vs-live-data): dated 2026-08-15 analysis reconciling the independent academic study [arXiv:2606.26028](https://arxiv.org/abs/2606.26028) (Xihan Xiong, Zelin Li, Wei Wei, Qin Wang, William Knottenbelt, Zhipeng Wang; Imperial College London; v1 2026-06-24, v2 2026-07-08; CC BY 4.0) against Onchain Agent Intel's live continuous index. States both predicate definitions explicitly side by side — the paper's "resolvable registration file AND >=1 declared live service endpoint" (frozen 2026-05-13, reporting 3% / 15% pass for Ethereum / Base) vs Onchain Agent Intel's stricter "endpoint 2xx AND at least one machine-readable capability surface (MCP / OpenAPI / .well-known)" (continuous). Direction and Base-vs-Ethereum ordering match under both predicates; see also the [all-Ethereum post](https://onchainagentintel.io/blog/why-we-went-all-ethereum-2026-07). Frames the commerce-backed cohort (our commerce-backed definition) as the architectural fix for the paper's central complaint that ReputationRegistry feedback is "rarely grounded in verifiable interactions." Not a rebuttal — a live continuation.
- [How to verify an ERC-8004 agent before you trust it](https://onchainagentintel.io/blog/how-to-verify-an-erc-8004-agent): evergreen how-to guide — the six on-chain and off-chain signals that separate a real, usable ERC-8004 AI agent from a paper registration (identity, ReputationRegistry feedback with Sybil-adjustment, x402 and settlement history, cross-chain presence, live reachability). Vendor-neutral checklist applicable on any chain.
- [Mapping an ERC-8004 agent's relationships on-chain](https://onchainagentintel.io/blog/mapping-agent-relationships-onchain): worked example on agent `base/19353` — what the free per-agent profile answers, what the free 200-node graph teaser shows, and what the paid `/v1/intel/peers/{agent_id}` and `/v1/intel/graph` endpoints add (the ten closest peer agents by capability overlap and the full inter-agent payment network in one call).
- [Most ERC-8004 agents are placeholders — the small slice that's actually live](https://onchainagentintel.io/blog/erc-8004-placeholders-and-fake-reputation): our ERC-8004 index on Base and Ethereum; only a few thousand of the registered total actually answer a request. Per-chain registered-vs-live figures pulled live from `/v1/public/stats`, with independent research (arXiv:2606.26028) corroborating the shape. Why the registry counts are the wrong number to shop by, and what a real trust filter looks like.
- [x402 Bazaar Market Map — live snapshot page](https://onchainagentintel.io/x402-bazaar-market-map): auto-refreshing reference page — Onchain Agent Intel's own daily snapshot of the public Coinbase CDP x402 Bazaar discovery catalog, cut by whether endpoints actually get called and paid in the last 30 days (the trust-filter tiers), broken down by network (CAIP-2 → human name), category, price distribution (USDC), top hosts by endpoint count, and the delta panel (adds / removes / price changes since the previous snapshot). Companion to the two long-form market posts above (the 2026-08-13 snapshot and the whole-catalog demand-concentration finding); this page tracks the same catalog on an ongoing daily cadence. Redistributable under CC BY 4.0 with attribution.
- [x402 Bazaar Market Map — JSON export](https://onchainagentintel.io/x402-bazaar-market-map.json): machine-readable JSON aggregate of the market-map dataset — cohort ladder, payer histogram, network / category / price / host summaries, delta panel — refreshed daily alongside the HTML reference page. Same source snapshot, same normalization, published under CC BY 4.0. Suitable for citation, re-hosting, or programmatic ingestion.
- [x402 Bazaar Market Map — CSV export](https://onchainagentintel.io/x402-bazaar-market-map.csv): CSV aggregate of the same market-map dataset — one row per payer-histogram bucket plus a TOTAL row, with the snapshot timestamp verbatim on every row. Refreshed daily alongside the JSON export and the HTML reference page. CC BY 4.0.
- [ERC-8004 Chain Liveness — daily per-chain live-endpoint & placeholder-name snapshot](https://onchainagentintel.io/erc-8004-chain-liveness): auto-refreshing reference page — Onchain Agent Intel's own daily snapshot of every ERC-8004 registration on Base + Ethereum, checked for whether it answers at its advertised HTTP endpoint (live rate) and whether its `name` field is empty or a placeholder shape (`Agent <id>` / `Agent #<id>`). Published as a top-level schema.org Dataset (Google Dataset Search-eligible) with CC BY 4.0 license and resolvable JSON + CSV distributions. Independently corroborated by [arXiv:2606.26028](https://arxiv.org/abs/2606.26028) (Imperial College, CC BY 4.0), which measures a similar predicate on a frozen 2026-05-13 snapshot and reports 3% / 15% pass for Ethereum / Base. Sibling dataset to [/erc-8004-agent-card-signpost-gap](https://onchainagentintel.io/erc-8004-agent-card-signpost-gap); together they measure the two independent axes (endpoint reachability + agent-card walkability) an autonomous discovery client must clear.
- [ERC-8004 Chain Liveness — JSON export](https://onchainagentintel.io/erc-8004-chain-liveness.json): machine-readable JSON aggregate of the chain-liveness dataset — envelope (publisher, license, canonical URL, snapshot timestamp) + totals (indexed / live / placeholder_name / named_and_live with derived percentages) + per-chain breakdown + one row per indexed agent (`chain`, `agent_id`, `registry_uri`, `endpoint_url`, `endpoint_status`, `is_live`, `is_placeholder_name`, `checked_at`). Refreshed daily alongside the HTML reference page. CC BY 4.0.
- [ERC-8004 Chain Liveness — CSV export](https://onchainagentintel.io/erc-8004-chain-liveness.csv): CSV aggregate of the same chain-liveness dataset — one row per indexed agent plus a TOTAL row that self-checks against the aggregate JSON. Refreshed daily alongside the JSON export and the HTML reference page. CC BY 4.0.
- [ERC-8004 Reviews From Paid Jobs — Base + Ethereum (commerce-backed reputation)](https://onchainagentintel.io/commerce-backed-agents-dataset): auto-refreshing reference page — Onchain Agent Intel's own snapshot of every ERC-8004 `ReputationRegistry` feedback event on Base + Ethereum that ties back to a real paid Virtuals ACP job or allowlisted ERC-8183 hook contract under the published commerce-backed definition. Published as a top-level schema.org Dataset (Google Dataset Search-eligible) with CC BY 4.0 license and resolvable JSON + CSV distributions. Sibling dataset to [/erc-8004-chain-liveness](https://onchainagentintel.io/erc-8004-chain-liveness) and [/erc-8004-agent-card-signpost-gap](https://onchainagentintel.io/erc-8004-agent-card-signpost-gap); together they measure the three independent axes (paid-job review attestation + endpoint reachability + agent-card walkability) a counterparty-quality filter needs. Same canonical predicate the `/v1/public/stats.commerce_backed_agents` surface and the free per-agent teaser use, so the dataset numbers cannot drift from the paid product.
- [ERC-8004 Commerce-Backed Reputation — JSON export](https://onchainagentintel.io/commerce-backed-agents-dataset.json): machine-readable JSON aggregate of the commerce-backed dataset — envelope (publisher, license, canonical URL, snapshot timestamp) + totals (all non-revoked feedback rows / commerce-backed rows / non-commerce remainder / unique-agent counts with derived percentages) + per-chain breakdown + one row per commerce-backed feedback event (`chain`, `agent_id`, `registry_uri`, `client_address`, `feedback_index`, `hook_type`, `endpoint`, `feedback_uri`, `score_norm`, `tag1`, `tag2`, `block_number`, `tx_hash`, `log_index`, `created_at`). Refreshed daily alongside the HTML reference page. CC BY 4.0.
- [ERC-8004 Commerce-Backed Reputation — CSV export](https://onchainagentintel.io/commerce-backed-agents-dataset.csv): CSV aggregate of the same commerce-backed dataset — one row per commerce-backed feedback event plus a TOTAL row that self-checks against the aggregate JSON. Refreshed daily alongside the JSON export and the HTML reference page. CC BY 4.0.

## Research citing our own data — additional citable datasets

- [ERC-8004 registry endpoint-gap dataset (N=577, 2026-09-15)](https://onchainagentintel.io/data/erc-8004-registry-endpoint-gap-2026-09): a cross-chain sample of 577 ERC-8004 identity records drawn from the 8004scan.io public index snapshot across five mainnet chains on 2026-09-15, temporally stratified. Pre-registered categorical claim C2 (`capabilities` AND `agent_url` null on 100% of the sample, 577 / 577) HELD against the 8004scan index and was not borne out at source: a source check of 43 rows against the metadata URLs the index itself captured and found 4 of the 18 resolvable live cards carry a non-empty `capabilities` field. The 100% null column is an index-side artefact of an explorer that stopped re-parsing agent cards on 2026-06-06, not a property of the underlying agents. Two other pre-registered claims (C1 endpoint_less_pct >= 60; C3 is_endpoint_verified true on 0 records) FAILED and are published alongside. Published as a top-level schema.org Dataset with CC BY 4.0 license and resolvable JSON + CSV + JSON-LD distributions.
- [ERC-8004 registry endpoint-gap dataset — CSV export](https://onchainagentintel.io/data/erc-8004-registry-endpoint-gap-2026-09/dataset.csv): 577 data rows plus header, 26 columns per row. CC BY 4.0.
- [ERC-8004 registry endpoint-gap dataset — JSON audit trail](https://onchainagentintel.io/data/erc-8004-registry-endpoint-gap-2026-09/summary.json): audit trail — full request URLs, forged cursor payloads, per-stratum age spread, budget used, pre-pinned claims and results. CC BY 4.0.
- [ERC-8004 registry endpoint-gap dataset — JSON-LD metadata](https://onchainagentintel.io/data/erc-8004-registry-endpoint-gap-2026-09/dataset.jsonld): schema.org Dataset metadata block, also inlined into the /data/ page for Google Dataset Search. CC BY 4.0.
- [Writing companion — 8004scan hasn't re-parsed a single ERC-8004 agent card since 2026-06-06](https://onchainagentintel.io/writing/erc-8004-registry-endpoint-gap-2026-09): editorial writing page for the endpoint-gap dataset — publishes both the 8004scan null-column measurement (577 / 577) and the falsifying source-verification pass that established the column is an index-side artefact rather than a property of the underlying agents.
- [ERC-8004 IdentityRegistry census — Ethereum, Base, Arbitrum (2026-09-17)](https://onchainagentintel.io/data/erc-8004-four-chain-census-2026-09): whole-history registration + distinct-sender census across Ethereum mainnet, Base, and Arbitrum One, checkpointed at 2026-09-17 with per-chain block anchors. Three full-census rows carry 100% `tx.from` coverage (Ethereum, Base, Arbitrum). Six rows total; three are in current scope. Public third-party totals disagree with each other by orders of magnitude in this window; this dataset publishes its own per-chain rows and its own method rather than adopting any of them. Seven named claim traps are enforced in the schema and re-flagged in `GRADE.md`. Published as a top-level schema.org Dataset with CC BY 4.0 license and resolvable CSV + JSON + JSON-LD distributions.
- [ERC-8004 three-chain census — CSV export](https://onchainagentintel.io/data/erc-8004-four-chain-census-2026-09/dataset.csv): 6 data rows plus header, 20 columns per row (chain, block anchor, distinct-sender count, coverage, Wilson CI95 bounds, caveats). Three rows (Ethereum, Base, Arbitrum). CC BY 4.0.
- [ERC-8004 three-chain census — JSON-LD metadata](https://onchainagentintel.io/data/erc-8004-four-chain-census-2026-09/dataset.jsonld): schema.org Dataset metadata block, also inlined into the /data/ page for Google Dataset Search. CC BY 4.0.
- [ERC-8004 three-chain census — DATA-DICTIONARY](https://onchainagentintel.io/data/erc-8004-four-chain-census-2026-09/DATA-DICTIONARY.md): column-by-column reader's map (20 columns) for the census dataset. CC BY 4.0.
- [Coinbase x402 Bazaar discovery census (2026-10-05)](https://onchainagentintel.io/data/x402-bazaar-census-2026-10): full-pagination snapshot of every resource listed in the public Coinbase CDP x402 Bazaar discovery endpoint at 2026-10-05T17:13Z (34,587 listings across 2,122 hosts, 173 pages, `truncated=false`). Headline findings: one host (`market.datapackvibe.com`) holds 38.98% of all listings; top-10 hosts hold 57.91%; HHI on hosts = 0.1571; 99.47% of listings accept at least one mainnet; the modal network is Base mainnet (`eip155:8453`); the median listed price is $0.01 USD per call with p90 = $0.03 and p99 = $0.945; 83.55% of priced listings cost ≤ $0.01; only 32 of 34,587 listings are MCP-type (0.093%). Includes a 100-URL seeded random liveness probe (40% HTTP 402 / 47% HTTP 405 / 1% HTTP 200 / 5% other 4xx / 7% net error). Our own `onchainagentintel.io` endpoints are explicitly disclosed as NOT present in the index as of the fetch window. Published as a top-level schema.org Dataset with CC BY 4.0 license and resolvable CSV + JSON summary + JSON-LD distributions. Source endpoint: `https://api.cdp.coinbase.com/platform/v2/x402/discovery/resources`.
- [Coinbase x402 Bazaar discovery census — CSV export](https://onchainagentintel.io/data/x402-bazaar-census-2026-10/dataset.csv): one row per listing, 21 columns per row (host, path, type, dominant network, scheme set, min USD price, mainnet-only flag, lastUpdated, curated flag, x402 version). 34,587 data rows plus header. CC BY 4.0.
- [Coinbase x402 Bazaar discovery census — summary JSON](https://onchainagentintel.io/data/x402-bazaar-census-2026-10/summary.json): aggregate metrics as a single JSON object — totals, host concentration (top 20 hosts, HHI, top-1 / top-10 share), network split (mainnet vs testnet accepts, top 20 networks), scheme usage, price distribution (percentiles, ≤$0.01 / ≤$0.10 / ≤$1 cuts), field completeness, recency bucketing, and the liveness-probe status breakdown. CC BY 4.0.
- [Coinbase x402 Bazaar discovery census — JSON-LD metadata](https://onchainagentintel.io/data/x402-bazaar-census-2026-10/dataset.jsonld): schema.org Dataset metadata block, also inlined into the /data/ page for Google Dataset Search. CC BY 4.0.
- [Coinbase x402 Bazaar discovery census — METHOD](https://onchainagentintel.io/data/x402-bazaar-census-2026-10/METHOD.md): fetch window, pagination, price-inference and testnet-classification rules, liveness-probe design, limitations, and explicit disclosure that our own endpoints are not currently listed in the Bazaar. CC BY 4.0.
- [Coinbase x402 Bazaar discovery census — DATA-DICTIONARY](https://onchainagentintel.io/data/x402-bazaar-census-2026-10/DATA-DICTIONARY.md): column-by-column reader's map (21 columns) plus the aggregate fields in summary.json. CC BY 4.0.
- [Coinbase x402 Bazaar discovery census — weekly delta MANIFEST](https://onchainagentintel.io/data/x402-bazaar-census-2026-10/weekly/MANIFEST.json): index of recurring weekly recomputations against the same source endpoint as the 2026-10-05 baseline. Each snapshot is written to a dated subpath (`weekly/<date>/summary.json` + `weekly/<date>/delta.json`) and never mutates in place; the 2026-10-05 fixed baseline sha256 is pinned in every `delta.json`. Latest week 2026-10-14 reports 34,412 listings (−175 vs. baseline), top-1 host share 28.72% (−10.26pp), HHI on hosts 0.08771 (−0.069403), week-on-week churn 51.78% (8,950 added, 8,865 removed vs. the 2026-10-07 recomputation at 34,648 items), plus a new `quality_summary` block exposing facilitator-reported per-resource `l30DaysUniquePayers` / `l30DaysTotalCalls` / `lastCalledAt` with Coinbase's own caveat verbatim ("facilitator-reported, NOT filtered for self-purchases or test wallets — catalog accounting, not endorsement"). CC BY 4.0.
- [Dataset license — schema & changelog (v1.0.0)](https://onchainagentintel.io/datasets/licence-schema): stable semver-versioned schema contract for the $19/mo dataset license — the weekly Parquet (CSV fallback) bulk extract of the free CC BY 4.0 [chain-liveness](https://onchainagentintel.io/erc-8004-chain-liveness) and [commerce-backed](https://onchainagentintel.io/commerce-backed-agents-dataset) datasets. Columns, types, nullability, and primary keys for both tables; 30-day dual-emit window on breaking changes; dated changelog section. The licensed extract columns are a strict subset of the free datasets — paid-tier evaluation fields (trust_verdict, Sybil-adjusted reputation with tag_breakdown, payment totals, owner handle strings, co-ownership graph, smart-account fields) are intentionally NOT shipped and are sold separately via the Pro subscription and per-call x402. Illustrative free-tier sample rows only; no paid data on the page. JSON Schema at `https://api.onchainagentintel.io/schema/dataset-licence/v1.0.0.json`. Published as schema.org Dataset with embedded Offer; schema page itself CC BY 4.0.
- [Narrative findings — We indexed all 34,587 x402 Bazaar resources. 39% come from one host, ~1% return 200.](https://onchainagentintel.io/blog/x402-bazaar-census-findings-2026-10): companion write-up to the 2026-10-05 x402 Bazaar discovery census. Walks through the full-pagination methodology (173 pages, `truncated=false`) and six findings cited with named `summary.json` keys: host concentration (one host at 38.98%, top-10 at 57.91%, HHI 0.1571), seeded 100-URL liveness probe (1% HTTP 200 / 40% HTTP 402 / 47% HTTP 405 / 5% other 4xx / 7% net error), price distribution (median $0.01, p99 $0.945001, 83.55% of priced listings at ≤ $0.01), network split (99.47% mainnet-accepting; Base dominant with 35,534 of 59,803 accepts), resource type (only 32 of 34,587 MCP-type — 0.093%), and recency (58.4% refreshed in the last 7 days). Explicit own-disclosure that zero of the 34,587 listings resolve to `onchainagentintel.io` as of the fetch window (`summary.json.ours_listed.count = 0`). Includes a "what this means for an agent choosing x402 services" layer-separation argument (listing ≠ liveness ≠ delivery) and a single soft CTA to the paid `/v1/intel/*` API + the free `/mcp` server. CC BY 4.0. Published as schema.org BlogPosting with isBasedOn linked to the Dataset.
- [ERC-8004 Arbitrum One cohort census (public RPC) — 2026-10-07](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10): first public-RPC-reproducible census of the ERC-8004 IdentityRegistry and ReputationRegistry on Arbitrum One at 2026-10-07T18:29:18Z. Full-history scan from the registry deploy block (428,895,443) to head (512,646,051) — 83,750,609 blocks, 9 windows, 100% coverage, 228 RPC calls on `https://arb1.arbitrum.io/rpc` (fallback `arbitrum.gateway.tenderly.co`), zero Alchemy credits, zero on-chain transactions. **1,604** distinct agent IDs from **1,000** distinct owners (top-1 owner `0xad7d4d39…ee721` holds 273 mints / **17.02%**; top-10 together hold **31.61%**). **973** `NewFeedback` events on the ReputationRegistry from **36** distinct clients to **108** distinct feedback-receiving agents (full-cohort full-history count — not an extrapolation). A 200-of-1,604 evenly-spaced tokenURI sample reports 99% carry a tokenURI, 22% resolve to parseable JSON, 20% carry service/capability/x402/payment signal, and **0/200 returned HTTP 2xx/3xx/4xx on a single-GET probe** because many Arbitrum agents publish ENS/`.eth`/`.arb`/`.agent` names in their endpoint field rather than HTTP(S) URLs — the 0% is a HTTP-reachability measurement, NOT an agent-liveness rate. Prior freeze (Alchemy archive, 2026-09-17): 1,502 registrations — **+6.79%** over ~20 days. Deploy block seeded from that freeze (public node is non-archive); deploy-block match verified. Source envelope SHA-256 `2ce26378ad36a7ded274f9f60ccd6b360b0239e574f5bceac39961260ba1554b`. **Arbitrum is MEASURED here, NOT indexed by the paid `/v1/intel/*` API, which covers Base and Ethereum.** Published as a top-level schema.org Dataset with CC BY 4.0 license and resolvable CSV + JSON + JSON-LD distributions.
- [ERC-8004 Arbitrum One cohort census — CSV export](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10/dataset.csv): 24 data rows plus header, 4 columns (metric, value, unit, note). Every published aggregate as a row — chain, chain_id, RPC endpoint, registry addresses, deploy block, head block, scan range, distinct agent IDs, distinct owners, top-1 and top-10 share, feedback events, distinct feedback agents and clients, tokenURI sample stats, prior-freeze registrations, delta, envelope SHA-256. CC BY 4.0.
- [ERC-8004 Arbitrum One cohort census — summary JSON](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10/summary.json): aggregate metrics as a single JSON object — source envelope pointers, measurement window (chain id, RPC endpoints, archive-ness, deploy/head block, scan coverage), registry contracts, identity registrations (distinct agents + distinct owners + top-10 owner table), reputation feedback totals, tokenURI sample (sample size, with-uri/resolved/signal/http-live rates, extrapolated-to-cohort figures, liveness caveat, top endpoint-host clusters), prior-measurement delta, coverage disclosure, and seven stated limitations. CC BY 4.0.
- [ERC-8004 Arbitrum One cohort census — JSON-LD metadata](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10/dataset.jsonld): schema.org Dataset metadata block, also inlined into the /data/ page for Google Dataset Search. CC BY 4.0.
- [ERC-8004 Arbitrum One cohort census — METHOD](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10/METHOD.md): block windows, sample strategy, probe design, archive-vs-full-node notes, delta vs. the 2026-09-17 four-chain census, and the explicit list of what the dataset does NOT publish (no cohort-wide Arbitrum live rate, no Base/Arbitrum ratio, no Arbitrum trust scores). CC BY 4.0.
- [ERC-8004 Arbitrum One cohort census — DATA-DICTIONARY](https://onchainagentintel.io/data/arbitrum-erc-8004-census-2026-10/DATA-DICTIONARY.md): column-by-column reader's map for `dataset.csv` (25 rows, 4 columns) and the top-level sections of `summary.json`. CC BY 4.0.
- [ERC-8004 Reputation Census — Ethereum vs Base (2026-10-09)](https://onchainagentintel.io/blog/erc8004-reputation-census-2026-10): dated per-chain census of the ERC-8004 `ReputationRegistry` (`0x8004BAa17C55a88189AE136b182e5fdA19dE9b63`) on Base and Ethereum mainnet, snapshot 2026-10-09 at 22:04 UTC. Base indexed through block 52,396,050 (2026-10-09 21:30 UTC); Ethereum mainnet indexed through block 26,142,739 (2026-10-07 19:55 UTC) — ETH indexer runs on a longer cadence. **570,562** `NewFeedback` events across **32,238** agents and **14,189** distinct client wallets (**549,229** / **21,333** by chain). Five headline numbers published inline: (1) Base is almost 26× the event volume of Ethereum from an 8× bigger agent set; (2) **93.7%** of Ethereum feedback and **19.2%** of Base feedback is submitted by **18 spray wallets** (pre-registered threshold ≥ 100 agents reviewed per wallet); (3) **BOTCOIN** (Base agent 25975) alone holds **71.2%** of Base feedback (390,890 events) from just **78** unique client wallets; (4) Ethereum's top-10 by raw count is **one 1,566-agent `IdentityMD #<n>` NFT batch collection** under a single owner, holding **51.0%** of Ethereum feedback; (5) after the 2026-10-10 per-agent payment attribution recompute (payments are now credited to the agent that declares its own payTo, not inherited from the owner wallet), the Transact-ready set across Base + Ethereum fell sharply from a previously-reported 1,046 (918 Base + 128 Ethereum under Rules C/D/F) to a live value published at `/v1/public/stats.readiness_buckets.Transact-ready` — nearly all prior badges were earning on inherited owner-treasury lines. Per-chain table with 7-day / 30-day feedback volumes, non-spray client counts, and readiness buckets. Top-10 Base and Ethereum tables published in both the raw-count view and the Sybil-aware unique-clients view. Full method inline (what counts as feedback, Rule C non-self predicate, spray threshold, block ranges). Companion to the 2026-10-09 readiness-scoring tightening. Aggregate CSV distribution at `/data/erc8004-reputation-census-2026-10.csv` (70 data rows, 8 columns: section, chain, rank, agent_id, display_name, value, unit, note; public agent IDs only, no client wallet addresses). Published as schema.org BlogPosting + Dataset (license, temporalCoverage 2026-01-29/2026-10-09, distribution link). CC BY 4.0.
- [ERC-8004 Reputation Census — aggregate CSV](https://onchainagentintel.io/data/erc8004-reputation-census-2026-10.csv): aggregate CSV distribution of the 2026-10-09 Base + Ethereum reputation census — 70 data rows across five sections (aggregate per-chain, readiness buckets, spray-wallet buckets, top-10 by raw feedback count, top-10 named by unique clients). 8 columns per row. Public agent IDs only; no client wallet addresses. CC BY 4.0.
- [Narrative companion — 1,604 ERC-8004 agents on Arbitrum, 108 with on-chain reputation. First public-RPC census.](https://onchainagentintel.io/blog/arbitrum-erc-8004-census-2026-10): companion write-up to the 2026-10-07 Arbitrum One ERC-8004 cohort census. Walks through the full-history methodology (nine 10M-block windows, 100% coverage via `arb1.arbitrum.io/rpc`, zero Alchemy credits) and five findings cited with named `summary.json` keys: 1,604 agent IDs from 1,000 distinct owners (1.604× mints-per-owner — wider operator base than Base's 3.02×); top-1 17.02% / top-10 31.61% concentration (real but not single-vendor); 108 agents with on-chain reputation feedback from 36 clients across 973 events (6.73% of registered cohort); 0/200 HTTP-reachable in the tokenURI sample explained by the ENS/`.agent`-named-endpoint pattern (not an agent-dead signal); +6.79% registration growth over the 2026-09-17 census. Explicit disclosure that Arbitrum is MEASURED here, NOT indexed by the paid `/v1/intel/*` API (which covers Base and Ethereum); this is not a coverage advertisement. Single soft CTA to the paid API + the free `/mcp` server. CC BY 4.0. Published as schema.org BlogPosting with isBasedOn linked to the Dataset.

## Optional

- [Atom feed](https://onchainagentintel.io/feed.xml): Atom 1.0 feed of every editorial post (State of ERC-8004 reports, worked examples, how-tos). Per-post real publication dates; feed `<updated>` is the newest entry's date so it is byte-stable until an editorial change lands. Suitable for newsletter ingestion, answer-engine polling, and aggregator subscription.
- [Robots](https://onchainagentintel.io/robots.txt): crawl policy.
- [Sitemap](https://onchainagentintel.io/sitemap.xml): sitemap for indexed human pages.