Reporting: what an agent did and paid

Reporting assembles a single verifiable report of everything an agent did and paid across the H-Series, joined by its identity on the public on-chain record. It answers “what has this agent registered, released, spent, and delivered,” in one timeline, with every line traceable to its source record on chain.

What it is

Given one identity (a CAIP-10 address like eip155:8453:0x... or hedera:mainnet:0.0.x, or an OIDC agent identity), the report fans out across the suite’s public records and joins them:

• Registrations (H-Index): services the agent registered.
• Credential releases and spend (H-Grant): gated releases, refusals with reasons, and the USD released.
• Execution receipts and requests (H-Seal): proof-of-execution receipts and the calls they cover.
• Deliveries (H-Relay): messages exchanged.
• Ring events (H-Pact), attestations (H-Cert), and merchant purchases (H-Agent): shown when present.

The result is a set of count cards, a headline “total verifiable spend,” a plain-language summary, and a timeline. The summary is deterministic (built from the counts and spend) and always present; an optional AI narration is available on demand. Every timeline entry expands to its full payload and links to the on-chain record, so nothing has to be taken on trust. The report also shows which product sources responded, and flags any that were unavailable or returned recent-only data.

How to use it

The report is at h-series.xr-utilities.com/report.

1. Enter an identity. Type a CAIP-10 or OIDC identity in the field and press “Build report,” or click one of the examples. A /report?identity=<id> link prefills and runs automatically.
2. Read the rollup. Count cards (registrations, releases, receipts, deliveries, and so on) sit above the total verifiable spend and the summary.
3. Filter the timeline. Narrow by product, type, result (released / refused / partial / error), date range, or free-text over the action and detail. Filtering is instant (client-side over the loaded events).
4. Verify a line. Expand any entry for its full untruncated payload and follow the on-chain reference to the source record.
5. Export. “Export CSV” or “JSON” writes exactly the currently filtered slice.
6. Narrate (optional). “AI summary” asks the server to narrate the report in prose. It is advisory and gated on a model key server-side; if it is off, the button says so and the deterministic summary still stands.
7. Go deeper (optional). Connect a wallet and sign an owner-read authorization to pull your complete indexed history across the suite on one signature, with refusal reasons and full detail beyond the recent chain replay.

Good to know

• The public report is a recent join. It reads recent on-chain records, so some sources may be flagged “recent only.” The deep (connected-wallet) report is the complete indexed history.
• Deep mode is EVM-only for now. The backend multi-scheme verification is live, but the non-EVM browser connectors are parked pending a fix, so the deep report currently accepts an EVM wallet.
• Spend is verifiable, not estimated. The headline sums releases, receipt-recorded call spend, and settled purchases. Non-USD currencies are listed separately rather than converted.
• Nothing here is signed or anchored by the report. It is a read-time join over records the products already anchored; the report itself never writes to chain.
• Results are cached briefly at the edge, so refreshing the same identity is cheap and does not hammer the mirror node.

Where to go next

• New to the suite? Read Start here for how the products compose into one lifecycle.
• For the estate-wide operator audit stream (all products, all accounts), see the SIEM guide.
• The spend lines come from grants and receipts. See the H-Grant guide, and set up an agent from the Console.