← All guides
Receipts · Zambo

What is an AI agent receipt?

An AI agent receipt is a public, persistent, independently checkable record of work done by an AI. It shows what was asked, what the agent actually did, and the observed result instead of asking someone to trust a screenshot or a claim.

The problem: AI work is trust-me work

Agents can say that they checked a price, audited a site, or scored a wallet, but the claim is often not portable proof. A screenshot proves only that an image exists; it does not prove what ran, when it ran, or whether the result was changed afterward.

A receipt makes the execution record inspectable at a stable public URL, with the request and observed result tied together by canonical bytes and a SHA-256 hash.

Anatomy of a verifiable receipt

A Zambo receipt carries the receipt ID, the SHA-256 of the canonical request bytes, a timestamp, the tool called and its arguments, the observed result, upstream evidence, and a Nostr anchor. These fields let another person or agent inspect the same record without relying on the original chat session.

See a real receipt

Open a verified live_price receipt →

The example above was checked at build time and returned HTTP 200 from the server-rendered /run/ receipt page.

Receipt vs. screenshot

Screenshots are trivially faked and disconnected from execution. A receipt binds canonical request bytes to the observed result with a hash anyone can recompute, so the public record can be checked independently.

Receipts and cross-AI continuity

AI work should outlive a single session or model. A receipt is the portable unit of cross-AI continuity: it gives the next agent a stable reference to what happened, what was observed, and what still needs to be done.

The model can change while the work remains available through its receipt and session trail.

FAQ

What is an AI agent receipt?

A public, persistent, independently checkable record of one specific AI task: the exact request, what the agent did, and the observed result — timestamped, SHA-256 hashed, served at a stable public URL.

How is a receipt different from a screenshot?

A screenshot can be faked in seconds and is disconnected from whatever actually ran. A receipt binds the canonical request bytes to the observed result with a SHA-256 hash, so anyone can recompute and check it.

What does a Zambo receipt contain?

A receipt ID, the SHA-256 of the canonical request bytes, a timestamp, the tool called and its arguments, the observed result, upstream evidence, and a Nostr anchor.

How do I verify a receipt?

Open its /run/ page: the receipt ID, SHA-256, canonical bytes, and observed result are all shown. Recompute the hash over the canonical bytes and compare.

Do receipts cost anything?

No. Every call on the free tier (20 calls per tool per day, no account) produces one. Paid tiers are Day Pass $1.49, $49/month, $399/year — re-verify against /pricing at build time; if changed, use the live values.

What does this have to do with cross-AI continuity?

Continuity means your work survives across sessions, handoffs, and model switches. The receipt is the portable proof that makes that possible — the work persists because the receipt persists.

Run your own free call

Connect the Zambo MCP endpoint, run a successful call, and receive your own verifiable receipt.

Run a free call at zambo.dev →

← Back to all guides · Read the cross-AI continuity guide →