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

# Retention and data

> How long fal Agent keeps prompts, generations, and chats, how organizations set a retention policy, and how external content is handled.

## Default retention

Self-serve accounts keep chats, prompts, and generations forever. You can delete a chat at any time from History, and you can remove a tile from a chat in the media rail.

## Organization retention policy

Enterprise accounts and teams under an organization can set a **Chat retention** period in **Settings → General**. The policy applies to the whole workspace.

| Options                                                                                                                                  |
| :--------------------------------------------------------------------------------------------------------------------------------------- |
| Forever, 1 hour, 2 hours, 6 hours, 12 hours, 1 day, 2 days, 3 days, 5 days, 1 week, 10 days, 2 weeks, 30 days, 90 days, 6 months, 1 year |

A team inherits the organization's policy unless it sets its own. The setting shows **Inherited from your organization** with a **Reset to org default** link when a team has overridden it. Only admins can change the policy.

### What the policy does

When a finite policy applies:

* Every upload and every model output is stamped with a matching lifetime on the fal CDN. The files expire.
* Every chat turn gets a matching expiry. Expired turns are hidden immediately and removed afterward.
* Request and response payloads for model runs are not archived.
* [Project memory](/docs/documentation/agent/projects/memory) distillation is turned off, because notes have no expiry.
* Project documents cannot be uploaded, because an indexed document has no expiry.

The policy applies to new turns from the moment it is set. It does not shorten the lifetime of existing files.

## What the agent shares with models

Model runs send your prompt and your attached or referenced media to the fal model endpoint, as with any fal API call. The agent's reasoning runs on a language model operated for fal. Chat content is not used to train models.

## External content

Content the agent fetches from the web, from a downloaded video, or from a connected app is treated as data. Instructions inside that content do not change the agent's behavior. The same applies to text inside uploaded documents.

## Sharing

A shared chat is visible to the audience you choose until the share expires or you revoke it. See [Sharing and forking](/docs/documentation/agent/chats/sharing).

## Delete data

* **Delete a chat** from History. This removes the chat and its turns.
* **Remove a tile** from the media rail. This removes it from the chat. The file stays in your Assets library until you delete it there.
* **Delete files in Assets** at [fal.ai/dashboard/assets](https://fal.ai/dashboard/assets).