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

# Large bulk changes

> How a change to thousands or millions of rows runs, and how to follow it

Every table has a bulk bar. Tick rows and the bar appears at the bottom of the
table with the changes you can make to them: set an owner, add a tag, dismiss,
restore, and so on.

Once every row on the page is ticked, the bar offers **Select all** with the
number of rows your filters match. Your change then applies to every one of
them, including the rows you have not scrolled to.

## Two ways a change runs

**Up to 25,000 rows, the change is made straight away.** You click, the rows
change, and the bar tells you how many changed.

**Above 25,000 rows, the change runs in the background.** The same is true when
the table cannot count its rows in advance: the bar then says **All matching
selected**. Before anything starts, Clustr asks you to confirm, with the number
of rows the change will touch. Then:

* The change starts and the selection clears. You can keep working, or leave
  the page.
* The bottom of the screen shows the change and how far it has got, for example
  **Setting owner, 34%**. Click it to see every running and recent change.
* When it finishes, Clustr tells you how many rows changed. If it fails, it
  tells you why and how far it got.

A large change makes exactly the same edits as a small one: the same rows
change, and each row's history records the change under your name. It just
makes them in batches, so a change to two million warm intros never has to fit
in one step.

## What happens to the rows

Clustr works out which rows match your filters when the change starts. Rows
that start to match afterwards (a new import, someone else's edit) are not part
of that change. Run it again to include them.

Each row is changed once. A change that stops part way (you cancel it, or
something goes wrong on our side) keeps the changes it already made. Running the
same change again is safe: rows that already have the owner, the tag or the
dismissal are left as they are.

## What each table can change

| Table | Changes |
| - | - |
| Warm intros | Set owner, add tag, dismiss, restore |
| Accounts | Set owner, add tag, set tier, dismiss, restore, add to targets, remove from targets, mark as clients, exclude |
| Connectors | Set owner, set tag, keep fresh, stop refreshing, remove |
| Raw signals | Export |

A few of them have limits of their own:

* **Exclude** stops before changing anything if the selection would take your
  exclusion list past its 500 companies.
* **Keep fresh** fills your plan's free refresh slots in list order and stops
  when they run out. The change then says how far it got.
* **Remove** deletes the connectors and their warm intros. It cannot be undone.
* **Restore** on warm intros also brings back an account that was dismissed
  with them. That account's other intros stay dismissed.

## Cancelling

Open the list at the bottom of the screen and click **Cancel** next to the
change. It stops after the batch it is writing. Changes already made stay made.

## Exports

An export of a large table also runs in the background. When it is done, the
list at the bottom of the screen offers **Download**. The file is kept for 7
days.

## What only runs on ticked rows

**Push to CRM** writes to your CRM one record at a time, within your CRM's
limits, so it only runs on rows you tick. A background change never writes to
your CRM. Make the change in Clustr, then push a smaller selection.

## From your AI assistant

The [MCP server](https://docs.clustr-ai.com/mcp-server.md) has the same lane: `start_bulk_job` starts a
background change from a table's filters, `list_bulk_jobs` and `get_bulk_job`
follow it, and `cancel_bulk_job` stops it.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.