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

# What Henry remembers

> Decisions, people, dates, and context. How memory is scoped, where it comes from, and how to correct or remove it.

Henry's memory is what turns a capable assistant into a teammate. It remembers so you do not have to repeat yourself, and it cites where each memory came from.

## What Henry keeps

* **Decisions**, with the date and the thread they happened in. "We decided on the 15th in #launch to push pricing to Q4."
* **People**: who does what, their preferences, and the dates attached to them.
* **How your team works**: tools, conventions, recurring processes.
* **Your preferences**: how you like things formatted, what to skip, what you always want included.
* **Task learnings**: quirks and lessons from scheduled runs, kept on each task.

When Henry recalls something, it says when it learned it and links the source. If two memories conflict, it trusts the newer one. Before relying on an old, time-sensitive fact, it verifies it.

## Where memory comes from

* **Conversations with you.** Anything you tell Henry, in any channel.
* **Channel capture.** In Slack and Teams channels Henry has been invited to, it reads the discussion so it can answer questions about team context. Turn this off per channel with `/henry capture off`, or remove what it learned with `/henry capture purge`. See [Slack](/user-guide/channels/slack).
* **Tool results.** Facts Henry reads from your connected tools while working.
* **Feedback.** Thumbs up and thumbs down, and corrections you make.

## Scope and privacy

* Memory is scoped to your **workspace**. Nothing crosses between workspaces.
* Some memory is **personal**: what you told Henry in a DM, your own preferences. Other members never see it.
* **Guests** from outside the workspace, such as Slack Connect users, get no memory and cannot read yours.
* Your data is never used to train models.

Detected secrets, such as API keys pasted in a channel, are redacted before storage.

## Correcting memory

Tell Henry. "That's wrong, the launch is the 22nd" or "forget my old phone number" updates its notes. Every change is recorded with a history, and where the **Skills** page is available you can review and revert any revision. See [Skills](/user-guide/memory/skills).

## Removing memory

* Per channel: `/henry capture purge` deletes everything learned from that channel.
* Per project: delete individual memory lines on the project page.
* Per workspace: admins can request a full data export and deletion. See [Security and compliance](/user-guide/admin/security-and-compliance).