Direct answer

Confirm agent task completion

To confirm agent task completion, you walk the task list against the record, not against the agent confidence. The agent says the task is complete. For each step it claims, there is either a verifiable receipt that passes the five checks, existence, identity, fingerprint, self-consistency, and chain, or there is not. Steps with passing receipts are confirmed as recorded. Steps without them are not complete yet, whatever the status line says. Run work through Zambo and the receipts exist from the start, checkable free by anyone asked to approve the result.

Confirm one call the way you would confirm a task: run the live demo, then confirm its receipt at zambo.dev/verify with the five checks below. The full method is on verify AI work.

Updated October 20, 2026.

The situation

Approval is the moment this question gets asked. An agent, yours, a teammate, or a contractor agent, reports a task complete: the leads were researched, the records were updated, the messages were sent, the files were produced. A person now has to approve the result, release a payment, close a ticket, or tell a client it is done. The agent claim is the only evidence on the table, and the approver did not watch the work happen, because not watching it was the point of delegating it.

So the approver has three bad options today: trust the summary, redo the work by hand to check it, or interrogate the chat transcript and hope the answers are evidence. Confirmation deserves a fourth option: a record that was fixed while the work ran, which the approver can check directly, step by step, without the agent, the operator, or the platform vouching for it. That record is what this page builds the confirmation habit on.

Why screenshots and chat logs fail as evidence

Status pages, screenshots, and chat logs all fail as confirmation for the same underlying reason: they are reports about the work, produced after the work, by parties with an interest in the approval. A screenshot of a completed task list confirms that a screen showed a completed task list. A log confirms that lines were written. Neither confirms that the call behind a step ran, returned what the step claims, or was not edited into a tidier story afterward.

A screenshot shows what a screen looked like at a moment someone chose. It can be cropped, staged, retaken after an edit, or taken from a different run entirely, and nothing inside the image lets a checker tell the difference. A chat log is weaker still: it is a transcript the operator controls, it can be edited or regenerated before it is handed over, and it records what the AI said, which is the claim, not the work. Logs kept by the tool itself are better, but they are usually private, mutable by whoever runs the system, and written for debugging rather than for a stranger to check.

Confirmation from receipts reverses the burden. The approver does not start from trust and look for reasons to doubt; they start from the task list and mark each step confirmed only when its receipt recomputes and its saved result matches the claim. Doubt becomes specific and cheap: not a general unease about AI, but a named step with no record, sent back to be run through the record. Everything else is confirmed and the approval can move.

The check, step by step

First, set the work up so there is something to check. Connect your AI client to the Zambo MCP at https://zambo.dev/mcp: one URL, no key, and it works with the clients and frameworks people already use, including ChatGPT, Claude, Claude Code, Cursor, Gemini, Grok, Codex, Copilot, VS Code, Windsurf, Cline, Zed, n8n, CrewAI, and LangChain. Zambo is the verifiable work layer for AI: 153 tools in total, 35 on the default surface, and every call made through it leaves a verifiable receipt as it runs, fixed at execution time instead of written up afterward. Then work through these steps:

Split the claim into steps. Take the task as the agent reported it and number its steps. Confirmation is bookkeeping at this point: step one needs the receipt for the call that did step one, and so on, in chain order, with no step confirmed on the strength of a neighboring step passing.
Run the work through the MCP. Each tool call saves its result and mints its receipt at the moment it runs. Work done outside the record leaves no receipt, so route the steps that matter through Zambo from the start.
Collect the receipts. Keep the receipt for each step with the deliverable, or gather the whole job into one handover link that carries the goal, the steps, the outputs, and the receipts together.
Open the free verifier. Go to zambo.dev/verify, enter the receipt, and run the check. It costs nothing and needs no signup, so the person checking never has to take the word of the person or the AI being checked.
Apply the five checks below to each step, then read the saved result itself instead of the summary written about it.
Decide from the record. Approve what the record confirms. Send back what it does not, by step number, to be run again through the MCP. Where a client or a manager is the approver, send the handover link instead of your own confirmation, so their approval rests on the same record yours did.

The five checks

Existence: a receipt exists for the call that supposedly did this part of the work. If there is no receipt, this step is unverified, whatever the summary says. Existence is the first gate when you confirm agent task completion, because a check cannot start from a story.

Identity: the receipt names the tool that ran, when it ran, and which run it belongs to. That stops a receipt from a different job, a different day, or a different tool being offered as evidence for this claim, and it ties each claim to one specific call.

Fingerprint: the digest recorded over the saved result recomputes from the result itself. If anyone changed the saved result after the fact, even by one character, the recomputed digest will not match the recorded one, and the check fails in the open instead of passing quietly.

Self-consistency: the parts of the receipt agree with each other and with the claim: the tool on the receipt is the tool the claim needs, the time falls inside the run window, and the saved result is the result the claim describes. A receipt can be intact and still be the wrong receipt for this claim.

Chain: the receipts for a multi-step job link in sequence, each one pointing back to the step before it. A missing step shows as a gap in the chain and a reordered step shows as a break, so the sequence itself is checkable, not just the individual calls taken one at a time.

The whole loop is the same one described on the family anchor page, verify AI work, and the receipt underneath it is defined on what is a verifiable execution receipt. For a finished run read as one ordered record, see the AI agent audit trail. Those three pages and this one describe one method from four angles: run the work so it leaves receipts, then check the receipts instead of the story.

Ways people mean this

Confirm agent task completion through an API: confirmation can be automated as a gate: a program walks the claimed steps, recomputes each receipt fingerprint, tests the chain, and returns confirmed or unconfirmed per step. Wire that gate in front of payment release or ticket closure and unconfirmed work stops moving by default.

Confirm agent task completion for developers: developers confirm at the level they already think in: tool calls, commands, and their results. The receipt for the test run and the receipt for the deploy step confirm what a status message cannot, and the chain shows nothing ran out of the order the task required.

Confirm agent task completion for clients: a client confirms from a handover link: the goal, the steps, the outputs, and a receipt per step, opened in a browser with no account. The client checks any receipt free and then accepts or flags from the same link, and the creator cannot accept their own handover, which is what makes the client confirmation worth having on record.

Confirm agent task completion with open source checking: AER-1 is an open IETF Internet-Draft, so confirmation does not depend on the producing platform staying online or staying honest. Any implementation of the draft can recompute the receipts, which makes a confirmation given today still checkable next year.

Confirm AI agent actions: actions confirm individually: the search action, the write action, the send action each resolve to a receipt naming the tool and the saved result. Confirm the actions and the task built from them is confirmed; skip the actions and you have only confirmed the summary.

Confirm AI work: the broader phrasing covers work that is not framed as an agent task: a draft produced, a dataset cleaned, an analysis run. The confirmation is identical, because the unit is identical: a claimed result, a call behind it, and a receipt that passes the five checks.

Confirm tool calls: to confirm tool calls, recompute each receipt fingerprint and read the saved result, paying closest attention to the calls with side effects: sends, writes, purchases prepared, records changed. Those are the calls where a confident summary and a different reality cost real money.

Confirm what my AI did: asked after the run, this is the proving question in confirmation clothing. The answer is the same record: receipts in chain order show what your AI did, and the steps missing from the chain are the answer to the follow-up question, what did it not do that it claimed.

What the check proves, and what it costs

Confirmation from a record proves the approver did their job. When a payment is later disputed, a client asks why a task was signed off, or a result turns out wrong, the confirmation trail shows exactly what was checked: which receipts, which saved results, which steps were sent back. That trail protects the approver as much as it disciplines the agent, and it turns approval from a vibe into a repeatable act that a second approver would perform the same way.

What it proves is narrow and useful: the saved result on the receipt was not changed after the fact, the call it records is the call the claim points to, and a multi-step job ran in the order its chain shows. That is the floor every argument about AI work currently lacks. Run the loop once on a real call first: the live demo runs one call and hands you its receipt, free, in about ten seconds.

The Day Pass is the fix for a day of unverifiable work

The tasks that most need confirmation are the ones nobody watched: the overnight run, the bulk update, the contractor agent job. A Day Pass makes the next such day confirmable for $0.99: route the day of work through the Zambo MCP so every call leaves its receipt, and approval at the end of the day is a walk down a checklist against a record, not an act of faith. One day, one dollar, and the completion claim arrives with its confirmation already attached.

For $0.99, a Day Pass covers 24 hours of recorded work through the Zambo MCP at https://zambo.dev/mcp. Ongoing use is the Zambo Pass at $29 a month or $290 a year, and $19 a month for $ZAMBO holders. Public handovers are free to send and free to open. Humans pay by PayPal or card through PayPal; agents pay USDC on Base through x402. Checking any receipt at zambo.dev/verify stays free either way.

Honest limits

Three edges, stated plainly. First, A verifiable receipt proves the saved result was not changed after the fact; it never proves the result was right, the work was well done, or the price was fair. That judgment stays with the checker, now made against a record that was actually checked. The record also only covers steps that ran through connected tools: work done entirely outside it has no receipt, and an honest record lists that work as unverified instead of smoothing it over. AER-1, the receipt format underneath, is an IETF Internet-Draft (draft-zambo-aer1-24), not a finalized standard. Second, confirmation confirms the record of a step, not its wisdom: a confirmed send can still have been a send that should not have happened, which is why side-effect steps deserve human review of their saved results, not just a passing fingerprint. Third, confirmation cannot reach backward: tasks completed before the work was recorded can only be checked by their outcomes, by hand.

Frequently asked questions

How do I confirm agent task completion?

Number the steps the agent claims it completed, then find the receipt for the call behind each step and run the five checks: existence, identity, fingerprint, self-consistency, and chain. Read the saved result for steps with side effects yourself. Every step whose receipt passes is confirmed as recorded. Any step with no receipt, a broken fingerprint, or a result that does not match the claim is not complete, and goes back by step number to be rerun through the record.

How do I confirm AI agent actions a contractor agent claims it took?

Ask for the record, not a report: the receipts for the run, or one handover link carrying the goal, steps, outputs, and receipts. Check a sample, or every side-effect action, free at zambo.dev/verify, with no account and no access to the contractor system. Actions that recompute and match are confirmed. Actions offered only as a summary, a screenshot, or a log export are unconfirmed, and payment terms can say exactly that.

Can a client confirm AI work without an account?

Yes. A client opens a handover link in a browser with no account, reads the steps and outputs, checks any receipt free at zambo.dev/verify, and then accepts the job or flags one named step from the same link. The creator cannot accept their own handover, so the client decision on the record is genuinely the client decision. That is what makes client confirmation usable later, next to an invoice or inside a dispute.

Does confirming tool calls prove the task was done well?

No. Confirming a tool call proves the call ran as recorded and its saved result was not changed afterward. Whether the call was the right call, and whether the overall task was done well, are judgments the approver makes against the confirmed record. Confirmation removes the factual argument about what happened. The quality argument remains, and it is a better argument once both sides are looking at the same checked facts.

How much does it cost to confirm a day of agent work?

Confirmation itself is free: checking any receipt at zambo.dev/verify needs no signup and costs the checker nothing, and the live demo is free. Recording the day of work costs $0.99 for a Day Pass covering 24 hours through the Zambo MCP at https://zambo.dev/mcp, 153 tools in total with 35 on the default surface. Ongoing use is the Zambo Pass at $29 a month or $290 a year, and $19 a month for $ZAMBO holders.