> ## Documentation Index
> Fetch the complete documentation index at: https://docs.prisme.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Data Retention

> Set how long conversations, documents, tasks, and artifacts are kept, within your plan cap

<Frame>
  <img src="https://mintcdn.com/prismeai/Y2OkYnuhIlMLI5jg/images/ai-governance-retention.png?fit=max&auto=format&n=Y2OkYnuhIlMLI5jg&q=85&s=f3388935504df88f5d4d452ae96a5b95" alt="Data Retention tab" width="3020" height="1618" data-path="images/ai-governance-retention.png" />
</Frame>

Data left in place forever is a liability, and the **Retention** tab is where you set how long each category of agent data lives before it is purged. Open it from **Agents Controls > Retention** in the Governance sidebar. This is an organization-level policy: it applies across every agent in the org unless an individual agent overrides it.

<Note>
  This data-retention policy governs operational agent data (conversations, documents, tasks, artifacts). It is distinct from the subscription-tier **audit log retention** covered in [Audit Logs](./audit-subscriptions), which controls how long administrative activity logs are kept.
</Note>

## Retention categories

You can set an independent retention period, in days, for each of four categories:

| Category          | What it covers                          |
| ----------------- | --------------------------------------- |
| **Conversations** | Chat threads between users and agents   |
| **Documents**     | Documents ingested into agent knowledge |
| **Tasks**         | Agent task records                      |
| **Artifacts**     | Files and outputs produced by agents    |

Each field takes a number of days. Leaving a field empty means no org-level value is set for that category, so resolution falls back to the platform default (shown as a placeholder in the field).

## The plan cap

Your subscription plan defines a maximum retention window (`plan_max_days`). The tab shows this cap as a badge, and each field is limited to it. If you enter a value above the cap, the save is rejected with a message telling you the maximum allowed. The backend enforces the same cap, so a value can never exceed what your plan permits.

## Three-stage resolution

The effective retention applied to any given piece of data is resolved through three stages, from most general to most specific:

1. **Platform default** : the baseline set by the platform, used when neither the org nor the agent specifies a value.
2. **Organization** : the value you set here in the Retention tab. It overrides the platform default for every agent in the org.
3. **Agent override** : a per-agent value that overrides the org value for that agent only.

The most specific stage that has a value wins. A category with no org value and no agent override falls back to the platform default.

## Agent-level overrides

Per-agent retention overrides are **not** set here. They are configured in the **Agent Creator**, per agent. When one or more agents in the org carry their own retention override, the Retention tab lists them read-only so you can see which agents deviate from the org policy without leaving Governance. To change an agent's override, open that agent in the Agent Creator.

## Saving changes

1. Go to **Agents Controls > Retention**.
2. Enter a number of days for each category you want to control, or leave a field empty to fall back to the platform default.
3. Keep each value at or below the plan cap shown in the badge.
4. Click **Save**.

<Info>
  Retention is enforced by the agent-factory backend. Changes apply to the whole organization and take effect for data purged after the policy is saved.
</Info>

## Next steps

<CardGroup cols={2}>
  <Card title="Model Governance" icon="robot" href="./model-governance">
    Control model access, agents, and knowledge storage
  </Card>

  <Card title="Audit Logs" icon="scroll" href="./audit-subscriptions">
    Subscription-tier audit log retention
  </Card>
</CardGroup>
