Before a model sees a request, and before a reply lands in the workspace, Kluff runs a checkpoint. Personal data is stripped, untrusted content cannot rewrite the task, and the response is checked.
From capability to everyday work
Before a model sees a request, and before a reply lands in the workspace, Kluff runs a checkpoint. Personal data is stripped, untrusted content cannot rewrite the task, and the response is checked.
“Prepare a support summary without exposing raw personal identifiers to the model.”
Select the support context available to the person asking.
Apply configured redaction before model processing.
Check the returned answer before using it in the next task.
Personal data removed before the model call
Personally identifiable information is redacted before it reaches a model. The company context stays in Kluff; the model call does not need the raw identifiers to do the work.
This is the default path, not an optional filter you remember to turn on for a sensitive thread.
Untrusted content cannot rewrite the task
Agents read tickets, emails, pages, and documents they did not write. Guardrails keep that untrusted content from becoming new instructions.
The task the person asked for stays the task. A pasted page or a connected inbox cannot quietly change what the agent is allowed to do.
Responses checked before they land
The reply is checked before it lands in the thread, the canvas, or the tool. That is the other half of the checkpoint: not only what goes into the model, but what comes back out.
Risky writes still wait for a person. Validation is not a substitute for approvals; it is the filter on the way through.
Before you try it
Work involving personal information, external documents, or content that needs careful review.
Identify the data types in your pilot and confirm the workspace’s redaction and model settings.
Review redaction settings and test representative inputs. Treat external content as context and check the result before taking action.
Connect your tools, ask your first question in a project, and watch memory, spend, and visibility stay in one place—not scattered across personal chats.
$15 / user · Unlimited AI
@kluff build a launch dashboard from HubSpot and the Q4 budget sheet. Refresh it every morning.

Done. Signups come from HubSpot, spend from the Q4 budget sheet. It refreshes at 7:00 AM.
v2 adds the channel split you asked for.

Can we flag any channel over $40 per trial?

Added. Paid social is flagged.
It sends from your Gmail, as you, with the draft attached. Nothing else leaves the project.
Ana Chen confirmed net 45 at 9:18 AM. Your rule: every customer email waits for you.