Aptoria
Features
Product
Resources
Company
Tools
Log in
See the demo
Help center
The AI agent
Your autonomy policy history
Every change to the agent's limits is recorded — what changed, who changed it, and why — and every past action carries the policy that was in force when it ran.
The short answer
Policy history records when your autonomy settings changed, who made the change, and what limits were in effect. Review it when an action needs context, and use the current policy screen to make new changes rather than treating old settings as a substitute for present approval.
Workflow at a glance
1
Every policy change is a record
2
Who changed it — you or the agent's suggestion
3
The policy behind every past action
4
Reviewing and walking a change back
Every policy change is a record
Your agent settings — the master switch, the per-feature toggles, the spend caps — are not just current state. Every time the policy changes, the change itself is logged: which fields changed, what they were before and after, when, and who or what made the change. The audit log merges these policy changes with the agent's actions into one chronological timeline, with a "Policy changes" count so you can see at a glance how often the dials have moved.
This is what makes your configuration accountable in the same way the agent's actions are. "The cap used to be lower" is not a memory — it is an entry, on the same tamper-evident, hash-chained record as everything else.
Who changed it — you or the agent's suggestion
Changes carry their source. A dial you moved yourself is recorded as your change. A change applied by accepting one of the agent's graduation suggestions is attributed to that suggestion, so the history distinguishes "the landlord widened this" from "the landlord accepted the agent's proposal to widen this." Either way a human made the decision — the agent never widens its own limits — but the record keeps the distinction honest.
Graduation suggestions themselves come with evidence: how many decisions of that kind you have approved, over how many days, at what approval rate, and the range of amounts observed. When you tap to apply one, the policy updates immediately and the change lands in the history with that provenance attached.
The policy behind every past action
Each autonomous decision also stores a snapshot of the policy in force at the moment it ran. So when you look at a past action in the activity feed or the audit log, you are not judging it against today's settings — you can see the exact thresholds the agent was operating under when it made that call.
Together, the two records answer the question that matters in any dispute or review: was the agent inside its limits, and whose limits were they? The policy history shows how the limits evolved and who moved them; the snapshots show which version applied to which action.
Reviewing and walking a change back
To review the history, open the audit log and filter the timeline — policy-change entries sit alongside agent actions, each listing the fields that changed. There is no separate "policy rollback" mechanism, and none is needed: to walk a change back, set the dial to its previous value on the Agent settings screen, and that change is recorded too. The history keeps both moves, so the record stays complete rather than being rewritten.
Related articles
The AI agent
Setting your autonomy thresholds
Turn features on one at a time, set the dollar and rate limits the agent works inside, and understand the soft-commit window that holds irreversible actions before they land.
The AI agent
Reviewing the agent activity log
The activity log shows every autonomous action the agent took, the plain-language reason behind it, whether it can be undone, and the exact policy it was operating under.
Security & compliance
Reading your audit trail
Every autonomous action is logged, explained, and — where possible — reversible. How to read the timeline, export it, and why it matters for disputes and compliance.
The AI agent
Using the emergency stop
The kill switch pauses all agent autonomy at once and voids anything sitting in its safety window, without touching your configuration. How to use it and how to resume.
Didn't answer your question?
Email support@aptoria.ai and a person will help — no ticket queue, replies within 24 hours on weekdays.
Email support
Back to Help center