Docs · How to

How to: read the Audit Log

A full walkthrough of the Audit Log — reading entries, and using it for security review, debugging and client reporting.

Trust in an AI agent comes from one thing above all: being able to see exactly what it did. The Audit Log is NIBWP’s answer — a complete, timestamped record of every tool call made against your site. This guide shows how to read it, what to look for, and how to use it for security reviews, debugging and client reporting.

TL;DR

The Audit Log records every ability call — which tool, which token, the arguments, the result, the timing. Use it to review what agents did, debug failures, and prove work to clients. It is always on; nothing an agent does is invisible.

Open the Audit Log

Go to NIBWP › Audit Log. You will see a reverse-chronological list of calls. Each row is one tool invocation, with enough detail to reconstruct exactly what happened.

What each entry tells you

FieldWhat it tells you
ToolWhich ability was called — e.g. wp-update-post
TokenWhich client/agent made the call
StatusSuccess, refused (out of scope), or error
TimingWhen it ran and how long it took
ResultWhat changed — the affected ID, the outcome

Three ways to use it

The same log serves three jobs:

  • Security review — after a bulk job, scan for anything unexpected. Refused calls show an agent reaching beyond its scope (and being stopped).
  • Debugging — when a workflow fails, the log shows the exact call and error, so you fix the cause, not the symptom.
  • Client reporting — export the log as a record of work done on a client site each month — your changelog and proof of work in one.
Check the log after every powerful runWhen you enable a write or destructive ability and run a bulk job, make reading the audit log the next step. It takes a minute and turns “I think it worked” into “I know exactly what changed.”

FAQ

Is logging always on?
Yes — every call is recorded automatically. You cannot accidentally run an agent “off the record.”
Can I see refused calls?
Yes — out-of-scope calls appear as refused, which is useful for spotting an agent (or a prompt) trying to do more than it should.
Can I export it?
The log is built for review and reporting — a complete record you can use to account for every change.

See everything your agents do

Every call, logged — review, debug and report with confidence.

Get NIBWP

Connect your first agent.

Follow the quickstart, or book a demo and we'll walk you through it.