Agent research receipts

An AgentResearchReceipt is a small JSON document an autonomous agent can publish after it researched something on Robinhood Chain. It says one thing: "I reached this conclusion from these facts, as of this time."

It is neutral and project-agnostic. It works the same for HoodLock, OrdoFi, Pons, HEY or any other project — HEY defines the shape and checks that the HEY evidence a receipt cites exists. HEY does not store receipts, count them, rank them, aggregate them into a consensus, or endorse a conclusion. A receipt is never a HEY signal and never an instruction to anyone else.

• JSON Schema (draft 2020-12): https://heyresearch.xyz/schemas/agent-research-receipt.v1.json
• Validator: POST https://heyresearch.xyz/api/receipts/validate (GET describes it)
• Source of truth in code: packages/domain/src/receipts/schema.ts (the Zod schema the validator runs and the JSON Schema it publishes, held equal by a test)

What a receipt carries

Evidence references are one of five kinds:

• hey_evidence — a HEY typed evidence id (ship:, signal:, abi:, impl:, lock:, source:, claim:, state:, narrative:, method:, sourcechange:), resolvable at /api/evidence/{id};
• hey_change_event — a ChangeEvent id from /api/changes, with the revision read;
• hey_snapshot — a project snapshot as read: the slug, its asOf, the scoringVersion;
• market_observation — a market reading: its provider and the time the provider observed it, and which figure (marketCap, fdv, price, liquidity, volume24h) — never a price target;
• external — any other public https:// URL and when the agent read it. HEY never fetches it.

Unknown top-level fields are refused, so there is no place to put a private chain-of-thought.

The agent's vocabulary, not HEY's

researchState.value and outcome.value are free strings the agent defines. Common ones:

HEY attaches no meaning to these and never turns one into a recommendation. An agent that uses a state such as HIGH RESEARCH CONVICTION defines it in researchState.definition.

Checking a receipt

curl -s https://heyresearch.xyz/api/receipts/validate \
  -H 'content-type: application/json' --data @receipt.json

The answer says whether the shape is valid (errors lists what is not, with paths), whether the subject project is published, and, for each cited HEY reference, one of:

• exists — a HEY evidence id or change event that stands. A change event cited with a revision is compared with the revision that stands now; cited without one, the answer carries reason: "revision_not_verified".
• revised — the change event stands, but HEY has revised it since the cited revision (reason: "current_revision:<n>"). A revision HEY never issued is not_found (no_such_revision).
• project_exists — a hey_snapshot: the project is published. The asOf the agent read is not verified (reason: "as_of_not_verified"; HEY does not rebuild the snapshot as it stood then), and scoringVersion says how the cited scoring version compares with the one HEY runs now: matches_current, differs_from_current or not_given.
• withdrawn (with HEY's reason), moved, not_found, invalid_id, or not_checked (another chain, a non-HEY reference, or past the limit of 25 HEY ids per receipt).

heyEvidenceStands (2026-09-28 review repair; it used to be true for a receipt HEY checked nothing in) is:

• true — every HEY reference was checked and stands (exists or project_exists);
• false — at least one checked HEY reference does not (missing, withdrawn, revised, moved or malformed);
• "partial" — every checked one stands, but some HEY references were not checked (heyEvidenceStandsReason: "some_not_checked");
• null — no HEY reference was checked at all: only external or market references, another chain, or none cited (heyEvidenceStandsReason: "nothing_checked").

The answer always carries stored: false and endorsement: false.

Limits: 64 KB of JSON per receipt, 30 checks a minute per client. Nothing the receipt names is fetched, whatever URL it contains.

Example: a project that is not HEY

The ids are illustrative; a real receipt cites ids it read from HEY. The outcome is the example agent's own, under its stated mandate.

{
  "type": "AgentResearchReceipt",
  "version": "1.0",
  "agentId": "https://agent.example.org/research-bot",
  "operator": "Example Research Collective",
  "mandate": "Monitor Robinhood Chain infrastructure projects for continued development; no allocation authority.",
  "subjectType": "project",
  "subject": { "chainId": 4663, "projectSlug": "hoodlock" },
  "createdAt": "2026-09-28T12:00:00Z",
  "evidenceCutoff": "2026-09-28T11:45:00Z",
  "thesisSummary": "The project shipped twice in the last 30 days and its lock contract is verified; its token market reading is older than HEY's freshness limit, so market context is set aside.",
  "claims": [
    { "statement": "A release was published on 2026-09-21.", "kind": "FACT", "evidence": [{ "kind": "hey_evidence", "id": "ship:3f1c2b1e-4d5a-4c6b-8e7f-9a0b1c2d3e4f" }] },
    { "statement": "HEY lists the activity status as SHIPPING under its published rule.", "kind": "DERIVED", "evidence": [{ "kind": "hey_snapshot", "project": "hoodlock", "asOf": "2026-09-28T11:40:00Z", "scoringVersion": "hbm-v16" }] },
    { "statement": "Development cadence looks steady to this agent.", "kind": "JUDGEMENT", "evidence": [] }
  ],
  "risks": [{ "statement": "The market reading is stale; liquidity today is not known from this evidence.", "evidence": [{ "kind": "market_observation", "source": "geckoterminal", "observedAt": "2026-09-26T08:00:00Z", "field": "liquidity" }] }],
  "unknowns": [{ "statement": "HEY holds no package registry source for this project.", "coverageDimension": "package" }],
  "researchState": { "value": "MONITOR", "definition": "Evidence of continued development; revisit on the next release or a coverage change.", "vocabulary": "https://agent.example.org/vocabulary/v1" },
  "outcome": { "value": "MONITOR", "decidedBy": "agent", "definition": "Keep watching; no action under this mandate." },
  "confidence": 0.6,
  "reviewConditions": [{ "condition": "A new release or a status change for the project.", "watch": "GET /api/changes?project=hoodlock&after=<cursor>" }],
  "sourceSystems": [{ "name": "HEY Research Lab", "url": "https://heyresearch.xyz", "apiVersion": "1" }]
}

Example: HEY itself, written the same way

{
  "type": "AgentResearchReceipt",
  "version": "1.0",
  "agentId": "did:web:agent.example.org",
  "mandate": "Assess research-infrastructure tokens on Robinhood Chain; may propose, never execute, an allocation.",
  "subjectType": "token",
  "subject": { "chainId": 4663, "projectSlug": "hey-research-lab", "tokenContract": "0x0000000000000000000000000000000000000001" },
  "createdAt": "2026-09-28T12:30:00Z",
  "evidenceCutoff": "2026-09-28T12:25:00Z",
  "thesisSummary": "HEY's own project shows recent ships and a hosted MCP in use; of its documented token utilities some are LIVE and others PLANNED per /api/hey/profile, and the valuation reading is an FDV, not a market cap.",
  "claims": [
    { "statement": "The profile lists Request Research as LIVE on this deployment.", "kind": "FACT", "evidence": [{ "kind": "external", "url": "https://heyresearch.xyz/api/hey/profile", "retrievedAt": "2026-09-28T12:25:00Z" }] },
    { "statement": "HEY recorded a release for its own project this week.", "kind": "FACT", "evidence": [{ "kind": "hey_change_event", "id": "ship:0b1c2d3e-4f5a-4b6c-8d7e-9f0a1b2c3d4e", "revision": 1 }] },
    { "statement": "Planned utilities are not counted as value by this agent.", "kind": "JUDGEMENT", "evidence": [] }
  ],
  "risks": [
    { "statement": "The valuation HEY holds is FDV; circulating supply differs.", "evidence": [{ "kind": "market_observation", "source": "dexscreener", "observedAt": "2026-09-28T12:00:00Z", "field": "fdv" }] },
    { "statement": "Liquidity depth is modest relative to this agent's minimum." }
  ],
  "unknowns": [{ "statement": "Whether the monthly treasury ledger has been published is not checked by the profile." }],
  "researchState": { "value": "RESEARCH_MORE", "definition": "Evidence is mixed or incomplete for this agent's mandate." },
  "outcome": { "value": "RESEARCH_MORE", "decidedBy": "agent", "definition": "Collect another month of evidence before any proposal." },
  "confidence": null,
  "reviewConditions": [{ "condition": "The first monthly treasury ledger entry, or a utility moving from PLANNED to LIVE.", "watch": "GET /api/hey/profile" }],
  "sourceSystems": [{ "name": "HEY Research Lab", "url": "https://heyresearch.xyz", "apiVersion": "1" }]
}

What was deliberately not built

• No storage and no aggregate. A surface that counted receipts or showed "agent research states" would only be honest with genuinely independent agents and operators behind it. None exist yet, and HEY will not create agent identities, run clones of one strategy, or present anything as consensus. If independent agents publish signed receipts later, a neutral aggregate (participants, states as each agent defined them, evidence cutoffs) is a founder decision.
• No execution. HEY provides research, evidence, receipts and monitoring. Any trade is a separately authorised decision of an external agent under its own risk policy; HEY runs no execution network and no copy-trade loop.
• No self-preference. A receipt about HEY is checked exactly like a receipt about any project, and nothing about receipts reaches activity status, Build Momentum, the Discovery Gap or the Builder Radar.