Published 2026-10-08 · CC BY 4.0 · Compilation of dated per-chain cohort reads
How many ERC-8004 agents are there? Per-chain adoption stats
Every public ERC-8004 adoption number sitting in a news story, slide
deck, or AI answer traces back to a scan of one of the on-chain
IdentityRegistry contracts. The specific contract — same CREATE2
address 0x8004A169FB4a3325136EB29fA0ceB6D2e539a432 across
every EVM chain where ERC-8004 is live — is the one authoritative
source. This page compiles the per-chain mint counts, distinct-owner
counts, and on-chain reputation-feedback counts from two full-rescan
datasets we have already published, and links the CSV / JSON behind each
number. No estimates, no extrapolations, no 8004scan derivative counts.
Headline numbers as of 2026-09-17 (Base, Ethereum) and 2026-10-07 (Arbitrum One). Each number below is a named field in one of the two source datasets — links alongside.
Headline numbers — per-chain registry mints and owners
| Chain | Registry mints | Distinct owners | Top-1 share | Top-10 share | As of |
|---|---|---|---|---|---|
| Base | 88,500 | 29,289 | 11.58% | 27.21% | 2026-09-17 |
| Ethereum mainnet | 50,820 | 11,352 | 4.92% | 18.25% | 2026-09-17 |
| Arbitrum One | 1,604 | 1,000 | 17.02% | 31.61% | 2026-10-07 |
Source: all three rows are named fields in
dataset.json
(dataset page);
the Arbitrum row is also carried in
summary.json
at the newer 2026-10-07 checkpoint
(dataset page).
On-chain reputation feedback — agents receiving ERC-8004 NewFeedback events
The ERC-8004 ReputationRegistry at
0x8004BAa17C55a88189AE136b182e5fdA19dE9b63 is a separate
contract from the IdentityRegistry. It receives a
NewFeedback(agentId, clientAddress, scoreUri, ...) event
every time a caller posts a reputation row for a registered agent. The
contract is permissionless — any address can post feedback for
any agent — so the row count is a signposting metric, not a
Sybil-resistant trust score. See
Can you trust an
ERC-8004 reputation score? for the live-measured gap between the
unfiltered feedback count and the commerce-backed subset.
| Chain | Agents with ≥1 feedback row | Feedback events | Distinct clients | As of |
|---|---|---|---|---|
| Base | 29,128 | 277,710 | (not measured) | live |
| Arbitrum One | 108 | 973 | 36 | 2026-10-07 |
| Ethereum mainnet | 0 | 0 | 0 | live |
Source: Base and Ethereum rows are the
agents_with_reputation_by_chain and
feedback_events_by_chain fields of the live
/v1/public/stats
endpoint (free, no key); Arbitrum row is the
reputation_feedback block of
summary.json.
Of the Base cohort, 44 agents carry
commerce-backed reputation under the predicate at
smartcontractauditpro/commerce_backed.py — the full
cohort is indexable at
/commerce-backed-agents.
Metadata quality — how many registered agents publish a resolvable card
A registry mint is cheap. Publishing a resolvable
tokenURI is slightly less cheap. Serving a live
agent-card.json at the URL that tokenURI
resolves to is work. The gap between those three is where almost all
of the "ERC-8004 looks huge but nothing happens" criticism lives.
| Chain | Non-empty tokenURI | % of registered | Notes |
|---|---|---|---|
| Ethereum mainnet | 30,502 | 60.9% | of 50,071 registry mints at the 2026-08-17 snapshot |
| Arbitrum One (sample) | 198 / 200 | 99.0% | evenly-spaced 200-of-1,604 sample; metadata resolves for 44 (22.0%) |
Source: Ethereum row is
ethereum_registry_non_empty_tokenuri /
ethereum_registry_mints on
/v1/public/stats;
Arbitrum row is tokenuri_sample.with_token_uri /
sample_n on the Arbitrum
summary.json.
Resolvable-metadata rate on Arbitrum is low because many agents publish
ENS, .eth or .agent names rather than HTTPS
URLs — a single-GET HTTP probe cannot resolve those. The
stronger live signal on Arbitrum is the 108-agent reputation-feedback
count in the table above.
Solana — separate registry, same question
ERC-8004 is an Ethereum standard, but a non-ERC on-chain agent
registry exists on Solana. We measured it once alongside the EVM
cohort for comparison. It is not part of the paid
/v1/intel/* index.
| Chain | Registrations | Distinct creators | Top-1 share | Top-10 share | As of |
|---|---|---|---|---|---|
| Solana | 1,477 | 185 | 22.1% | 78.7% | 2026-09-02 |
Source: the solana_census block on
/v1/public/stats;
raw census snapshot at
solana_8004_census.json.
Methodology
-
Base, Ethereum, Arbitrum (2026-09-17 census).
Full rescan of the ERC-8004 IdentityRegistry at
0x8004A169FB4a3325136EB29fA0ceB6D2e539a432, decodingTransfer(from=0x0, to=OWNER, tokenId)logs from the verified deploy block to a per-chain checkpoint around 2026-09-17T04:31Z. Registrations = decoded mint count; distinct owners = uniquetoaddresses in the decoded logs. Full dataset: /data/erc-8004-four-chain-census-2026-09 (CSV + JSON + JSON-LD, CC BY 4.0). -
Arbitrum One (2026-10-07 full-history census).
Full rescan on
arb1.arbitrum.io/rpc(fallbackarbitrum.gateway.tenderly.co) from the ERC-8004 deploy block 428,895,443 to head 512,646,051 — 83,750,609 blocks, 228 RPC calls, 0 errors, zero Alchemy credits. The ReputationRegistryNewFeedbackscan runs over the same block range on0x8004BAa17C55a88189AE136b182e5fdA19dE9b63. Full dataset: /data/arbitrum-erc-8004-census-2026-10. Narrative: 1,604 ERC-8004 agents on Arbitrum, 108 with on-chain reputation. -
Base + Ethereum reputation + metadata (live).
Rows pulled from the
/v1/public/statsendpoint (free, no key, no payment). Fields cited:agents_with_reputation_by_chain,feedback_events_by_chain,ethereum_registry_mints,ethereum_registry_non_empty_tokenuri,ethereum_registry_snapshot_at,commerce_backed_agents,solana_census. -
What the numbers measure. A "registry mint" is a
Transfer-from-zero event on the IdentityRegistry: it is a count of
registrations, not a count of agents that currently serve a live
agent-card.jsonat their advertised URL. A "distinct owner" is a unique minting address: an operator that minted many agents counts once. "Agents with reputation feedback" is a count of distinctagentIds that appear on the receiving side of at least oneNewFeedbackevent; because the contract accepts feedback from any caller, this is a signposting metric, not a Sybil-adjusted trust signal (see Can you trust an ERC-8004 reputation score?). We report the raw count and separately maintain the commerce-backed subset at /commerce-backed-agents. -
What this page is not. This is not a
count of live agents at their advertised endpoints. The
liveness-at-endpoint number for the cohort On-Chain Agent Intel
indexes (Base + Ethereum) is in
live_by_chainon/v1/public/statsand is a materially smaller number than the registration count: on the measurement date, Base shows ~9% endpoint-liveness and Ethereum shows ~47%, net ~13% across the indexed cohort. Those numbers are updated as agents come up or go down; the mint counts on this page do not change retrospectively.
Datasets — CSV / JSON downloads
-
Cross-chain census (Base, Ethereum, Arbitrum):
dataset.csv·dataset.json·dataset.jsonld(schema.orgDataset) · dataset page -
Arbitrum One full-history census:
dataset.csv·summary.json·dataset.jsonld· dataset page -
Live cohort stats (Base + Ethereum, measured today):
api.onchainagentintel.io/v1/public/stats -
Solana census snapshot (comparison, not indexed):
solana_8004_census.json
Related reading
- 1,604 ERC-8004 agents on Arbitrum, 108 with on-chain reputation — first public-RPC census
- ERC-8183 Agentic Commerce — joining the DRAFT ERC to live ERC-8004 reputation
- 8004scan's 100% capabilities-null column is an index-side artefact
- Can you trust an ERC-8004 reputation score?
- Commerce-backed cohort — ERC-8004 agents with reputation tied to a real paid job
Licensed CC BY 4.0 — reuse the numbers, cite the page. Cite as: Onchain Agent Intel (2026). How many ERC-8004 agents are there? Per-chain adoption stats. https://onchainagentintel.io/research/erc8004-adoption-stats.