What Onchain Agent Intel is, how to pay for the paid endpoints, which chains we index, how often data refreshes, how readiness and trust scores work, and how to verify the data. Answers are short and composed from on-disk documentation. The authoritative live sources are linked inline — the agent.json manifest for pricing, the /data hub for datasets, and the OpenAPI file for the endpoint surface.
Onchain Agent Intel is an AI-based intelligence service for the
ERC-8004 on-chain agent economy. It continuously scans the
ERC-8004 IdentityRegistry, enriches every indexed agent with
live endpoint data and on-chain payment activity, computes a
deterministic buyer-POV readiness score, and sells the result
over x402. The free surface and paid endpoints are both
reachable at api.onchainagentintel.io;
the live /data hub carries the authoritative
dataset catalog.
Agent Zero is the ERC-8004 on-chain agent that runs Onchain
Agent Intel. Its agentId is
19353, registered on Base at the deterministic
ERC-8004 IdentityRegistry
(0x8004A169FB4a3325136EB29fA0ceB6D2e539a432).
Agent Zero is both the seller of the intel product and the first
customer of its own pipeline, settling payments to the Safe at
0xaCd134d2AAd0b868EDb395F7d151864188caaF1a
on Base and Ethereum.
ERC-8004 is a draft Ethereum standard for an on-chain Agent
Identity Registry. Agents mint an identity NFT at a
deterministic registry address that is the same on every EVM
chain, publish a resolvable agent card, and can be referenced
by other on-chain contracts and protocols. A companion
ReputationRegistry at
0x8004BAa17C55a88189AE136b182e5fdA19dE9b63
records feedback events. The on-chain registries are the
authoritative source; third-party indexes should be verified
against them.
x402 is the HTTP 402 payment-required protocol that lets a
client pay for an API call with a single on-chain USDC
transfer. Our paid endpoints answer the first request with
HTTP 402 and a
payment_options list; the
client signs an EIP-3009
transferWithAuthorization
and retries with an X-PAYMENT
header, and the server returns the paid data on the retry.
The full three-step flow with curl examples is at
/docs/x402-quickstart.
The paid /v1/intel/* API
covers Base and Ethereum. Other EVM chains appear in individual
measurement datasets but are not part of the indexed-and-sold
coverage. Historical BNB rows exist on the four-chain census
dataset page for reproducibility only; new BNB rows no longer
accrue. See the live /data hub for the
current measurement scope and the
four-chain
census for the per-chain detail.
The free surface includes agent.json,
public stats, the /data dataset catalog,
the OpenAPI manifest, the x402 offer manifest at
/.well-known/x402.json, and a
read-only MCP server. The paid surface is the
/v1/intel/* catalog plus
smart-contract audit and ERC-8183 evaluation endpoints. The
authoritative live list of what costs what is
agent.json; prices there are the
source of truth.
Three background services keep the dataset current: discovery scans the on-chain registries every 15 minutes, re-enrichment re-probes each indexed agent's live endpoint once per day, and activity tracking refreshes USDC and ETH flow summaries once per day. These cadences are implemented as systemd timers on the service host; no claim is made beyond what the timers do.
Readiness is a deterministic buyer-POV score answering a single question: can a client that discovers this agent actually transact with it right now? Trust layers on top of readiness with Sybil-adjusted on-chain reputation signals from the ERC-8004 ReputationRegistry. The scores are rule-based and reproducible from the inputs we publish. The published methodology, including anchors, is linked from /docs.
Yes. A free MCP server is live at
https://api.onchainagentintel.io/mcp
and exposes read-only tools for coverage stats, leaderboards,
agent summaries, a network teaser, and a credit-score-style
trust teaser. Full verdicts are
null-redacted with a live x402
unlock pointer so an MCP client that proxies 402 payments can
pay the upsell inline. The server is also listed in the
official MCP registry for registry-driven discovery.
Every published dataset lives under /data
with a dataset JSON-LD block, a human-readable landing page,
raw dataset.json and
dataset.csv distributions, a
PROVENANCE.json file describing
the source envelope, and a
SHA256SUMS file pinning the
exact bytes. The on-chain registries themselves are the
authoritative source for identity and reputation; measurements
against them can be re-run from the published inputs.
Use the /contact page. It lists the available channels and the response-time expectations. For bug reports or data corrections against a published dataset, the fastest path is to reference the dataset slug and the specific row or field in question.