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

# Commands

> Save and reuse prompts to automate repetitive analyses.

Commands are reusable prompts that save you time on repetitive work. Instead of typing out the same detailed question each time, create a Command once and run it whenever you need it.

<Note>
  Need to manage commands outside the UI? See [Manage Commands via API](../guides/manage-commands-via-api).
</Note>

## Why Use Commands?

Many analyses follow the same pattern:

* Preparing for a QBR with a specific account
* Running weekly portfolio health checks
* Assessing churn risk for renewal accounts
* Generating onboarding status reports

Without Commands, you'd type out (or copy-paste) the same detailed prompt every time. With Commands, you save the prompt once and run it with a single click.

## Creating a Command

1. Click **Commands** in the sidebar
2. Click **New Command**
3. Give your Command a name (e.g., "QBR Prep")
4. Write the prompt you want to reuse
5. Add any variables for dynamic content (optional)
6. Click **Save**

### Using Variables

Variables let you customize a Command each time you run it. Wrap variable names in double curly braces:

```
Prepare a QBR summary for {{account_name}} including:
- Health score trends over the last 6 months
- Product usage highlights
- Open support tickets
- Expansion opportunities
```

When you run this Command, Humm will ask you to fill in `account_name` before executing.

## Running a Command

1. Click **Commands** in the sidebar
2. Find the Command you want to run
3. Click **Run**
4. Fill in any variables if prompted
5. Humm creates a new thread with the Command's prompt

The result is a regular thread, so you can ask follow-up questions and dig deeper.

If you have a long command library, use the search bar at the top of the Commands page to filter by command name or description.

## Command Ideas

<CardGroup cols={2}>
  <Card title="QBR Preparation">
    Generate a comprehensive account review with health metrics, usage trends, and talking points.
  </Card>

  <Card title="Weekly Portfolio Review">
    Summarize health changes, at-risk accounts, and expansion opportunities across your book of business.
  </Card>

  <Card title="Churn Risk Assessment">
    Analyze warning signs for accounts approaching renewal, with recommended actions.
  </Card>

  <Card title="New Customer Onboarding">
    Track onboarding progress, milestone completion, and early adoption metrics.
  </Card>
</CardGroup>

## Sharing Commands

Commands can be personal or shared with your team:

* **Personal Commands** — Only you can see and run them
* **Team Commands** — Anyone on your team can use them

To share a Command, open it and click **Share with Team**. Shared Commands appear in everyone's Command library.

## Save a Thread as a Command

At the end of a productive thread, choose **Save as Command**. Humm extracts the question pattern and auto-detects variable placeholders (`{{variable}}`), so you don't have to build the command from scratch.

This is the fastest way to create a command — let a real conversation prove out the analysis first, then save it for reuse.

## Using @ References in Commands

Use `@` references to pin a command to specific tables or fields. This keeps results consistent even as your schema evolves.

```
Show weekly active users from @[table:events] grouped by @[column:events.plan_tier]
```

See [Threads — Using @ References](/features/threads#using--references) for the full syntax.

## Editing and Deleting Commands

Open any Command to edit its name, prompt, or variables. Changes apply immediately—the next time anyone runs the Command, they'll get the updated version.

### Edit with the Assistant

If you want help tightening a prompt instead of rewriting it by hand:

1. Open the Command
2. Start an edit conversation with the assistant
3. Describe the change you want
4. Review the proposed diff
5. Approve the update when it looks right

The assistant works against a draft first, so you can inspect the proposed changes before they go live.

### Review History and Restore an Older Version

Every saved update creates a history snapshot. Open **History** on a command to:

* Review earlier versions
* Compare an older snapshot against the current version
* Restore a previous version if you need to roll back a prompt change

Restoring an older version makes that snapshot the live command again. Humm also records the restore in version history so you keep a clear audit trail.

To delete a Command, open it and click **Delete**. Deleted Commands can't be recovered, so make sure you don't need it anymore.

## Tips for Effective Commands

<AccordionGroup>
  <Accordion title="Be detailed in your prompts">
    Commands work best when the prompt is specific. Include exactly what sections, metrics, and format you want in the output.
  </Accordion>

  <Accordion title="Use clear variable names">
    Name variables descriptively: `{{account_name}}` is clearer than `{{x}}`. This helps when running the Command later.
  </Accordion>

  <Accordion title="Test before sharing">
    Run your Command a few times with different inputs to make sure it produces good results before sharing with your team.
  </Accordion>

  <Accordion title="Start from a successful thread">
    If you asked a great question in a thread and want to reuse it, copy the prompt into a new Command. You already know it works.
  </Accordion>
</AccordionGroup>