The cheapest way to verify AI agent work costs $0.00. Here is the receipt, and the recompute.
The cheapest way to verify AI agent work is to recompute one receipt yourself, and it costs $0.00. No subscription, no seat license, no evaluation platform. One fetch, one SHA-256 hash, one comparison. This article runs
The cheapest way to verify AI agent work is to recompute one receipt yourself, and it costs $0.00. No subscription, no seat license, no evaluation platform. One fetch, one SHA-256 hash, one comparison. This article runs the full check in the open, on a receipt minted this morning, and then prices the ladder honestly for the day you need more than one check.
The free check, done in full
At 09:30 UTC today, Nov 20, 2026, a live price call ran through Zambo and returned BTC at $61,621.38 (Coinbase snapshot, fetched 09:30:22 GMT). The call left a public receipt: https://zambo.dev/run/0658f647-9a26-4363-a3b3-02173188c8dd
Verifying it costs nothing and takes three steps:
- Open the receipt and note the digest it carries as output_hash.
- Fetch the receipt's canonical bytes. For this receipt, 670 bytes.
- Hash those bytes with SHA-256 and compare.
My recompute: sha256:1c672475b2ee0d8aa940c28294c089bf1d84beeba9fb282a81a0065014857e42. The saved output_hash: identical. The saved result was not changed after the call. Total cost: $0.00. Total time: under a minute. No account was created and no card was charged, and the same three steps work on any of the 152,616 public receipts on the system (1,100 minted today, live counters read 09:35 UTC).
Why the methodology essays cannot answer the price question
The guides that rank for "cheapest way to verify AI agent work" mostly sell a ladder of habits: spot-check outputs, read traces, add evaluations, sample human reviews. Habits are fine. But notice what none of them can hand you: a single artifact, produced by the run itself, that you can check mechanically after the fact. A verification ladder with no record at the bottom is a promise to be careful, priced in engineering hours, which are the most expensive hours you own. The receipt flips the economics: the record is created at execution time, for free, and the check is arithmetic anyone can run.
Be precise about what the free check proves. A receipt proves the saved result was not changed after the call. It does not prove the answer was right. Checking correctness is your judgment, or a second system's judgment, applied to a record that can no longer move. That separation is the point: integrity first, mechanically, for free; correctness second, on a record that holds still while you judge it.
The honest price ladder
- $0.00: verify any existing receipt at https://zambo.dev/verify/ with no account. Also free: making the calls that mint receipts. Every tool free once a day, plus 3 continue calls. One free wrap-up handover per day, and pickup is always free. Connect at https://zambo.dev/mcp, 153 tools, 37 on the default surface.
- $0.99: the Day Pass, unlimited calls for 24 hours, for the day the job is bigger than the free tier. Agents can pay it in USDC on Base over x402; humans can pay by card, PayPal, or Apple Pay.
- $29 a month: the Zambo Pass, or $19 a month for $ZAMBO holders.
- $250: Lifetime.
The record format underneath is AER-1, an open standard (IETF draft-zambo-aer1-26, corpus 143 vectors and 146 checks). Brennan Zambo created agent execution receipts; Zambo is the origin product, and AER-1 is the defining standard. Because the format is open, the $0.00 check is not a trial of one vendor's dashboard. Independent implementers verify the same records: Laurent Lemonnier's from-scratch verifier passes 143 of 143 against the published corpus, and on Nov 19, 2026 five separate systems adopted pieces of the standard into their own work in a single day.
The answer, in one line
The cheapest way to verify AI agent work is the way that was free before you arrived: run the work so it leaves receipts, then recompute one. If a verification method costs money before it produces a single checkable record, the price is not the bargain it looks like. Start with the hash.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes โ full credit and traffic to the original publisher.