Decision infrastructure for digital assets.
When the ChainChoice agent names a pick, the answer is saved as a receipt at a public URL: content-hashed, stamped with the build, model, prompt and tool set that produced it, and stored with the market figures the answer was given. A record you can attach to an investment-committee memo or a counterparty risk file.
The pick and the next rows of the order ChainChoice's ranking engine returns for the question: the category's published ranking re-run with the stated country and the questionnaire answers read from the question, keeping only rows confirmed eligible for that country. Each row carries its position for those answers beside its published rank and ChainChoice Score. No fee or cost figure: the agent states none.
The published category and country, the questionnaire answers the engine ran, how many rows it set apart and the first of them with the engine's reason, and for the rows shown their published rank, ChainChoice Score and score components, with the methodology version and the date the category's evidence was read. Yield fields are filled only from a dated source. Before the answer is shown or stored, a sentence that orders the rows differently, names a row the engine set apart, or prints a score out of 100 is removed, and the trace lists each removal with the rule it broke. Figures in the answer text are checked against these values and the market snapshot; a figure that matches neither and names no recognised source is logged, and removed only where enforcement is switched on.
Methodology version, agent build SHA, model id, prompt version and tool-stack hash, recorded when the receipt is written. They identify the code, model, prompt and tools that produced the answer. They do not make it reproducible: the answer is model-written text, and nothing re-runs a receipt.
sha256 over the canonical block of everything the receipt states. prev_hash is the hash of the previous receipt in the same anonymous session; a receipt with no session starts no chain. After the write, ChainChoice updates only fields outside the hash (a recorded outcome, moderation status, account binding); an edit to a hashed field no longer matches its hash.
The market figures relevant to the question — asset prices, Ethereum gas, the ETH staking rate — captured with a materiality threshold for each. When a receipt is opened, current values are compared with them and moves past the threshold are flagged. The comparison is computed on read and not stored; the receipt itself is not changed.
Not advice. Not a fiduciary product. Not custodial. We do not execute, hold assets, or take discretion. ChainChoice is the audit trail under whoever does — internal allocator, external advisor, treasury operator, compliance reviewer.
Receipts for agent answers that name a pick: a public URL, a content hash, the build and model stamp, a per-session hash chain, and a drift check on the market figures when the receipt is opened. The partner API returns the pick with its published rank and score, the engine's order for the question beside the published order, and the receipt URL.
A source archive (snapshots of cited URLs), an append-only governance log (approvals, exports, escalations) and successor receipts when a watched figure moves. The database has tables for all three; nothing writes to them yet. No date is set.
API/infrastructure tier — the receipt schema as a service. Embedded in advisor stacks, treasury management software, and audit toolchains so the artifact follows the decision wherever it goes.
We are pre-pilot. If you allocate to digital assets and want decision evidence at institutional grade — or if you operate audit infrastructure that should consume this schema — reach out. Email institutional@chainchoice.io with a one-line description of your use case.