A random-sample observational probe of ERC-8004 agent
tokenURI hop-H2 resolvability, measured on
2026-09-02T20:06:20Z. For each sampled agent we
issued one unauthenticated HTTP(S) request at hop H2 and placed
the response into one of nine pre-defined buckets.
Per-chain sample size n=600 on Base + Ethereum;
pooled sample size n=1200.
This measures schema shape and HTTP resolvability on
2026-09-02, not agent quality — see
METHOD.md.
No agent is scored, ranked, or bucketed by this dataset; the
usable_card_definition is a schema check pinned
before probing, not a ranking rule. For the current / live
picture of indexed agents, see /ecosystem.
Headline table
| Scope | Resolve rate | Usable-card rate | n |
|---|---|---|---|
| Base | 68.2% (64.3–71.8) | 56.3% (52.3–60.2) | 600 |
| Ethereum | 74.8% (71.2–78.1) | 71.3% (67.6–74.8) | 600 |
| Pooled | 71.5% (68.9–74.0) | 63.8% (61.1–66.5) | 1200 |
Rates are the point estimate with the 95% Wilson confidence interval in parentheses. Values are stored at full source-JSON precision in dataset.csv; the percentages above round for display only.
Buckets observed, by chain
| Bucket | Base | Ethereum | Pooled |
|---|---|---|---|
ok_json |
409 | 449 | 858 |
of which usable_card |
338 | 428 | 766 |
of which json_but_not_a_card |
71 | 21 | 92 |
http_4xx |
181 | 81 | 262 |
timeout |
9 | 1 | 10 |
dns_or_conn_fail |
0 | 69 | 69 |
bad_scheme_or_unparseable_uri |
1 | 0 | 1 |
Nine buckets were defined before probing; only the ones with at
least one observation are shown above. Zero-observation buckets:
non_json_body, empty_body,
http_5xx, ipfs_gateway_fail. The full
bucket list is in METHOD.md.
Usable-card definition (pinned before probing)
A response counts as a usable card only when all three pre-conditions hold:
requires:
- JSON object at top level
- non-empty top-level `name` string
- at least one of: url, endpoint, serviceEndpoint, serviceEndpoints,
services, capabilities, skills
pinned_before_probing: true
The rule was pinned before any probe was issued. Responses that
parsed as top-level JSON but failed one or more pre-conditions
are counted as json_but_not_a_card and are the
source of the near-miss shape examples below.
Top host, by bucket
DNS domain only — no per-agent data. Share is
top_host_n / bucket_n.
| Chain | Bucket | Top host | Share | n / bucket |
|---|---|---|---|---|
| Base | ok_json |
<data> |
25.9% | 106 / 409 |
| Base | http_4xx |
ipfs.io |
97.2% | 176 / 181 |
| Base | timeout |
termix-platform-prod.s3.ap-southeast-1.amazonaws.com |
88.9% | 8 / 9 |
| Base | bad_scheme_or_unparseable_uri |
<unparseable> |
100.0% | 1 / 1 |
| Ethereum | ok_json |
api.normies.art |
49.7% | 223 / 449 |
| Ethereum | http_4xx |
ipfs.io |
93.8% | 76 / 81 |
| Ethereum | dns_or_conn_fail |
api.freaks.one |
98.6% | 68 / 69 |
| Ethereum | timeout |
slops.app |
100.0% | 1 / 1 |
Near-miss shape examples
The raw JSON lists 3 near-miss shape examples per chain
(per_chain.<chain>.near_miss_shape_examples) —
a diagnostic listing of the top-level JSON keys present on an
ok_json response that failed the usable-card rule. See the
raw JSON for the full set.
Downloads
| File | Format | Notes |
|---|---|---|
| dataset.csv | text/csv |
6 data rows + header. Three scopes × two metrics. No rounding. |
| tokenuri_resolvability.json | application/json |
Raw observational probe payload: corpus metadata, pinned usable-card rule, 9-bucket definitions, per-chain + pooled aggregates with Wilson 95% CIs, top-host-by-bucket summaries (DNS domain only), 3 near-miss shape examples per chain. Already live. |
| dataset.jsonld | application/ld+json |
Schema.org Dataset JSON-LD mirror of the in-page block. |
| SHA256SUMS | text/plain |
BSD sha256sum output for the sibling files in this directory. |
| PROVENANCE.json | application/json |
Source task, source doc, probe timestamps, raw JSON URL, licence. |
Reading
| Document | Purpose |
|---|---|
| METHOD.md | Eligible-population build, sampling, probe parameters, pinned usable-card rule, 9-bucket definitions, Wilson CI note, and what this is NOT. |
| DATA-DICTIONARY.md | Field-by-field schema for the raw JSON (top-level + corpus + per_chain.<chain> + pooled) and the derived summary CSV. |
License and citation
Licensed CC BY 4.0 — reuse it, cite it. Cite as: On-Chain Agent Intel (2026). ERC-8004 tokenURI resolvability (hop H2), 2026-09-02. CC BY 4.0. https://onchainagentintel.io/data/tokenuri-resolvability-2026-09-02/.