← Home · Data hub · Blog

ERC-8004 Chain Liveness

Data as of 2026-09-07T08:09:08.976732+00:00 UTC · refreshed on every daily site build from the live ERC-8004 profile index (snapshot status: ok).

Every ERC-8004 registration on Base + Ethereum, checked for whether it answers at its advertised HTTP endpoint. Of 25,724 indexed agents, 6,187 (24.1%) return HTTP 200. A further 111 (0.431%) carry only a placeholder-shape name (Agent <id> or empty).

Per-chain breakdown

Indexed registrations, live-endpoint (HTTP 200) count and share, and placeholder-name (Agent <id> or empty) count and share. Every number derives from the same snapshot the /v1/public/stats surface reads.

Method

Source. The ~/.openclaw/erc8004_agent_profiles.sqlite index — the same profile store the /v1/public/stats surface and the signpost-gap dataset read — built by agent_activity_tracker.py and agent_reenrich.py against every ERC-8004 registration the discovery script has ever seen on Base + Ethereum.

Registry. ERC-8004 IdentityRegistry at 0x8004A169FB4a3325136EB29fA0ceB6D2e539a432 — identical CREATE2 address on Base (eip155:8453) and Ethereum mainnet (eip155:1). One row per ERC-8004 mint (Transfer from the zero address on the IdentityRegistry) checked against our periodic live-endpoint probe.

Live predicate. endpoint_status == 200 on the last probe of the URL the agent advertises as its primary service endpoint. This is the same predicate liveness.is_live uses for the paid /v1/public/stats surface, so the free counts and the paid product cannot drift. Necessary condition for a discovery client to reach the agent; not a sufficient one — a live endpoint may still lack a discoverable agent-card (measured by the sibling signpost-gap dataset) or fail a stricter capability probe.

Placeholder-name predicate. The profile's name field is either empty (null / empty string) or matches the discovery / SEO-fallback shape Agent <id> / Agent #<id>. This is the same predicate the internal audit_placeholder_names.py report uses. The placeholder-name rate published here is a floor on the human-attributable-name rate (agents whose token metadata carries a real name have already been resolved into name by the enrichment pipeline), not a ceiling on the working-agent rate — an agent with a real name can still be dead at the endpoint layer, and an agent with a placeholder name can still be live. The two axes are independent by construction; the CSV / JSON exports carry both as separate boolean columns.

Independent corroboration

Imperial College London's academic study arXiv:2606.26028 (Xiong, Li, Wei, Wang, Knottenbelt, Wang; v2 2026-07-08; CC BY 4.0) measures a similar predicate on a frozen 2026-05-13 snapshot and reports 3% (Ethereum), 4% (BSC), and 15% (Base) of registrations pass their trust bar — the paper's conclusion is that the ReputationRegistry “cannot function as a trust signal” in its current deployed state. The live-data reconciliation post states both predicate definitions side by side and shows direction and Base-vs-Ethereum ordering match under both.

Limitations

• Index ≠ whole chain. We index only ERC-8004 registrations the discovery script has already scanned; new mints appear in the next scan cycle, not instantly. A run-time drift of a few hundred rows is normal, and the exact numbers here move day to day.
• Sample vs population. Every figure on this page is a population figure over the full indexed set as of the snapshot timestamp. Any sample-of-N traversal claim on the sibling signpost-gap traversal audit is a sample figure and must not be blurred with the population figures published here.
• Endpoint churn. Live status is refreshed on the activity-tracker cadence (approximately every two days per agent), not on every build. checked_at per row captures when the last on-chain check ran; older checked_at values indicate stale rows.
• Placeholder-name is a floor, not a ceiling. See the Method section for the exact predicate definition. Broader claims about “most registrations are placeholders” that appear elsewhere in the ecosystem generally reflect the “not live at the endpoint layer” population, which is an entirely different measurement — use the Live column, not the Placeholder-name column, when comparing against those claims.

Cite this dataset

Headline finding. As of 2026-09-07T08:09:08.976732+00:00, 6,187 of 25,724 (24.1%) indexed ERC-8004 registrations on Base + Ethereum return HTTP 200 at their advertised endpoint. A further 111 (0.431%) carry only a placeholder-shape name (`Agent <id>` or empty) — a floor on the human-attributable-name rate, not a ceiling on the working-agent rate.

Citation. On-Chain Agent Intel. "ERC-8004 Chain Liveness — daily per-chain live-endpoint & placeholder-name snapshot." Snapshot 2026-09-07T08:09:08.976732+00:00. https://onchainagentintel.io/erc-8004-chain-liveness

BibTeX. Copy-paste for LaTeX / biblatex / Zotero:

@misc{ocai_erc_8004_chain_liveness,
  author       = {{On-Chain Agent Intel}},
  title        = {ERC-8004 Chain Liveness — daily per-chain live-endpoint \& placeholder-name snapshot},
  year         = {2026},
  month        = {09},
  howpublished = {\url{https://onchainagentintel.io/erc-8004-chain-liveness}},
  note         = {Snapshot: 2026-09-07T08:09:08.976732+00:00. License: CC BY 4.0.},
}

License. Published under CC BY 4.0 — free to reuse, including in commercial work, provided you attribute “On-Chain Agent Intel” and link back to https://onchainagentintel.io/erc-8004-chain-liveness.

Machine-readable exports. JSON · CSV — the same numbers rendered above, refreshed on every daily rebuild.

Also useful: All datasets · Signpost-gap dataset · Base chain brief · Ethereum chain brief · Commerce-backed cohort · Live-data reconciliation