Direct answer
Your boss needs an audit trail of what the agent did. Hand them one link.
The work product survives an audit; the trail does not. What the agent actually did, in what order, with which inputs, usually lives in a chat scroll mixed with three other jobs, and a screenshot of it convinces nobody. When the work runs on Zambo, the trail assembles itself while the work happens: every call inside one job is a step with a verifiable receipt, decisions included, and the wrapped job is the audit packet, one link that shows the whole thing in order, checkable by anyone it is handed to.
Updated November 2, 2026. First published November 2, 2026. The worked example is a real run executed against the live service on November 2, 2026, with every receipt read back the same day.
New to this? Start here, no code
Add the Zambo endpoint (zambo.dev/add) to your chat AI before the agent starts work, then paste this block.
The paste
Run this job as an audit trail on Zambo: 1. Connect to https://zambo.dev/mcp. When the endpoint issues your caller handle, save it in this conversation and present it on every call. 2. Open one job: pick a single job id for this work and keep every call inside it. Tool calls, decisions, and saves all land as steps in the one job, each with its own receipt. 3. Save decisions as steps too: when you choose a course of action, call memory_remember under the job id with the decision and the reason in under 500 characters. An audit trail that shows actions but not decisions is half a trail. 4. At the end, wrap the job with handover: a summary naming the job and its steps, a goal, and the auditor or manager as the recipient. The handover is the audit packet: one link, every step in order, every receipt checkable. 5. Hand over the link, not a summary of the link. Anyone who questions a step opens it and checks its receipt at https://zambo.dev/verify/ themselves.
Why agent work fails an audit today
The work product survives; the trail does not. The summary email exists, the deliverable exists, but what the agent actually did, in what order, with which inputs, lives in a chat scroll that mixes three jobs and a lunch order, if it survives at all.
Screenshots are not a trail. A screenshot shows a moment someone chose to capture, cropped the way they chose to crop it. An auditor cannot check it against anything, and both sides know it.
A trail has to assemble itself. Nobody will hand-log an agent's steps after the fact, and a log written afterwards is a story, not a trail. When the work runs on Zambo, the trail is the work: every call is a step with a verifiable receipt, saved as it happens, inside one job.
The pattern
Step zero: the operator and the auditor hold separate caller handles; the auditor's reading is a different identity acting on the record, which is what makes it an audit.
1. One job for the work: every call lands in the same job id, in order.
2. Decisions are steps: saved with memory_remember alongside the tool calls they steer.
3. Wrap the packet: one handover whose proof loop binds the complete step list and checks each step's receipt.
4. The auditor walks it: pickup is always free; the auditor reads each step from the record and checks receipts at the verify endpoint, then files the packet link.
The real run: one job, walked end to end by a second identity
Executed November 2, 2026, against the live endpoint, job uc-audit-2026-11-02. The operator ran three receipted steps: an ETH price check, $2,248.28, receipt e422411b; a saved decision naming that price step as the rate reference for a weekly vendor summary, receipt 5e140e8c; and a SOL cross-check, $91.57, receipt 50967f66. The operator wrapped the job: the audit packet (JSON twin here).
Then the part that makes it an audit instead of a claim: a second identity, the auditor, picked the packet up free, walked all 9 steps in order, and checked every step's receipt at the v2 verify endpoint: 9 of 9 returned verification_status verified, matching the packet's own receipt checks (checked 9, not checked 0). The auditor filed two notes into the same job, the walk and the completion: receipt fbe9da6e and receipt fb682e8f. The packet reads back status accepted. That link, plus the auditor's notes, is what goes to the boss: not a summary of what the agent probably did, but the thing itself, checkable by anyone it is handed to. For the wider frame, the audit trail article and the hub's AI agent audit trail page go deeper.
Audit this packet yourself
Open the audit packet, pick any step, and check its receipt at zambo.dev/verify. A receipt proves the saved result was not changed since it was saved, which is the property an audit trail exists for: the record of the work is the record as it was made, not as anyone later wished it read. It does not prove the operator chose well; it proves what was done, which is the question an audit asks. If a receipt from this run ever fails to verify, that is news: write to brennanzambo@zambo.dev.
What this costs
The free tier is per identity: 20 calls + 1 wrap-up free every day for each identity's own caller handle, and picking up in another AI is always free, so the auditor's reading never costs the operator anything. Counted from the run: the operator spent 3 counted calls plus the free wrap; the auditor spent 2 counted calls (its two filed notes) and its pickup was free. A team running audited agent work all day steps up per identity: a Day Pass is $0.99, unlimited for 24 hours, and the Zambo Pass is $29 for 30 days, one-time payments, never subscriptions. The trail is the product here: jobs run any other way leave nothing to audit. Start at zambo.dev/add.
What Zambo is not here
Zambo is not a compliance department and does not certify that a job was done well or legally. It keeps the step-by-step record, with receipts, that a manager, a client, or an auditor can walk themselves. Whether the trail satisfies a given regulator stays your organization's call; that the trail exists and has not moved does not.
Frequently asked questions
My boss needs an audit trail of everything my AI agent did. What do I hand them?
One link: the job's handover. Run the work on Zambo inside one job id and every call, tool call or saved decision, becomes a step with a verifiable receipt. Wrap the job and the handover is the packet: steps in order, each receipt checkable at zambo.dev/verify. On the run on this page, a second identity walked all 9 steps and verified 9 of 9 receipts.
Is a chat export an audit trail?
No. A chat export shows what was said, in the agent's own words, mixed with everything else in the scroll, and nothing in it can be checked against a record. The packet's steps are the calls themselves, each receipt proving its saved result has not changed since it was saved.
Does the trail prove the agent did good work?
No. A receipt proves the saved result was not changed since it was saved. The trail proves what was done, in order, which is the question an audit asks. Judging whether it was done well stays with the humans reading it.
What does an audited job cost?
The run on this page used 5 counted calls across two free allowances, the auditor's pickup free. At volume, the Day Pass ($0.99, unlimited for 24 hours) and the Zambo Pass ($29 for 30 days) apply per identity; both are one-time payments, never subscriptions.