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

# Personal Memories

> How to configure Unblocked to remember your preferences and apply them to future answers

Every person interacts with Unblocked differently. For example, a backend engineer wants code-level
answers with stack traces. A support lead needs copy-paste blocks for customer replies. A product
manager wants the customer's point of view, not the implementation details. Without memories, each of
them has to re-state those preferences every time they ask a question.

Personal Memories let Unblocked learn your preferences — role, answer style, toolchain, terminology,
even which documents to trust or ignore — and apply them automatically to every answer.

Where [Shared Memories](/unblocked-chat/shared-memories) remember facts about your *organization* and
shape what everyone is told, Personal Memories remember preferences about *you* and shape how you're
answered.

## What Unblocked remembers

Personal Memories go beyond answer length and language. These are the cases where they make
a noticeable difference:

* **Your role and context.** "I'm a support engineer on the billing team." Unblocked scopes answers
  to your domain and stops explaining things you already know.
* **Answer format and tone.** "Always give me a table comparing the options." Answers arrive in the
  shape you want without you asking for it each time.
* **Toolchain and environment.** "I only use Claude Code" or "give debug examples using gdb."
  Answers reference the tools you actually work in.
* **Terminology and aliases.** "When I say *ToDo* in Linear, I mean the *Unstarted (Todo)* status."
  Unblocked translates your shorthand into the right query.
* **Sources to leave out.** When there is nuance in a connected data source, you can specify "don't
  use the auto-generated draft docs in Google Drive as ground truth," for example. Unblocked then
  skips documents you know to be unreliable, so they can't mislead an answer.
* **Templates for recurring questions.** "Whenever I ask for the report, give me an up-to-date
  version of the chart above." A long, fiddly request becomes a one-liner.

## How it works

Unblocked learns a memory in one of two ways:

* **You tell Unblocked directly.** When [asking Unblocked a question](/unblocked-chat/asking-questions),
  state a preference outright and Unblocked saves it immediately — "I'm a software engineer," "always
  answer from the customer's point of view," "reply in Spanish."
* **Unblocked notices a pattern.** When you ask for the same kind of change repeatedly — shorter
  answers, replies reformatted as tables, a topic you keep returning to — Unblocked promotes it to a
  saved memory. A single request isn't enough; the preference has to show up more than once, so
  Unblocked learns habits rather than one-offs. Because these patterns are gathered in the
  background, memories learned this way can take up to a day to take effect.

Memories shape both how Unblocked answers you — tone, format, length, language, and focus — and how
it gathers context, including which documents it treats as reliable and how it turns your shorthand
into a search. What a memory never does is override what your connected sources actually say.

### Where it works

Personal Memories work only in **private, one-on-one conversations with Unblocked** — for example, a
direct chat in the web app, the Mac app, or a direct message with the Unblocked bot.

<Note>
  To keep personal preferences out of shared spaces, memories are switched off
  whenever a conversation isn't private.
</Note>

* **Shared conversations.** As soon as a conversation is [made public](/unblocked-chat/incognito-mode),
  or when someone else joins a conversation, Unblocked stops capturing or applying memories.
* **Public channels.** In a public Slack or Microsoft Teams channel, Personal Memories are never
  used, regardless of who is in the channel.

## Examples

### Match your toolchain

If your local agentic tool of choice is Claude Code, tell Unblocked once:

```
I only use Claude Code for all my work.
Remember that, so your answers are always for that environment.
```

Configuration help now arrives with only Claude Code in mind, instead of listing every editor and
agent Unblocked supports.

### Learn a new codebase

If you want answers that teach rather than just resolve:

```
I'm trying to learn the codebase. When you answer questions about it,
explain the underlying concepts too. Act like a staff engineer mentoring me.
```

### Scope answers to a temporary role

On-call for the week and fielding customer questions:

```
I'm on-call this week. When you answer my questions, always include
copy-and-paste blocks I can use to draft replies to customers.
```

When the rotation ends, ask Unblocked to forget it.

### Rule out an unreliable source

Some documents shouldn't be treated as fact — drafts, generated output, an archived wiki nobody
maintains:

```
The auto-generated draft docs in our Google Drive are under review and often wrong.
Never use them as input or ground truth.
```

Unblocked leaves those documents out when it gathers context for your answers.

### Turn a long request into a short one

Once you've walked Unblocked through a report you want regularly, save the shape of it:

```
Whenever I ask you for the report, give me an up-to-date chart like the one above.
```

From then on, "give me the report" is enough.

<Note>
  This example needs a database connected to Unblocked — see
  [Snowflake](/data-sources/snowflake).
</Note>

Keep this to a preference about the answer you want. When what you need is a procedure — the steps to
follow, the rules to apply when a certain kind of question comes up — write a [skill](/skills)
instead.

## Setup and management

### Managing your memories

You manage memories in conversation — just ask Unblocked in a private chat:

* **See what it knows:** "What do you remember about me?"
* **Add or change a preference:** "Remember that I prefer detailed, step-by-step answers," or "from
  now on, answer in French."
* **Correct something:** "Update my role — I'm a product manager now."
* **Remove a memory:** "Forget that I want answers in French."

### Turn Personal Memories on or off

Personal Memories are on by default once the feature is enabled for your organization.
You can opt out at any time:

<Steps>
  <Step title="Open your preferences">
    In the Unblocked web app, open **Settings** and select **Preferences** under **Personal Settings**.
  </Step>

  <Step title="Toggle Personal Memories">
    Find the **Personal Memories** toggle and turn it off to opt out, or on to opt back in.
  </Step>
</Steps>

When you turn Personal Memories off, Unblocked stops learning new memories and stops applying
existing ones to your answers. Your existing memories are kept, not deleted — if you turn the
feature back on later, they apply again.