Product

AI Guardrails

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

AI Guardrails, in practice.

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.

Use it for
Work involving personal information, external documents, or content that needs careful review.
Start with
Identify the data types in your pilot and confirm the workspace’s redaction and model settings.
Read the setup guide
Illustrative workflow Sample data
The request

“Prepare a support summary without exposing raw personal identifiers to the model.”

  1. Select the support context available to the person asking.

  2. Apply configured redaction before model processing.

  3. Check the returned answer before using it in the next task.

A request with explicit data boundaries

Example output
Input
Authorized support context
Protection
Configured redaction and model path
Output
A summary for the team to review

PII redaction

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.

Trust and data handling

Prompt injection protection

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.

Output validation

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

Questions for your evaluation.

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.

Unlimited inference.
Company memory.

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
Q4 launch› Launch dashboard
Sara Castillo

@kluff build a launch dashboard from HubSpot and the Q4 budget sheet. Refresh it every morning.

Kluff

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.

Maya Ruiz

Can we flag any channel over $40 per trial?

Kluff

Added. Paid social is flagged.

Q4 launch dashboard
Q4 launchHubSpot + Q4 budget sheet · refreshed 7:00 AM
Live
Signups1,284+18% vs last week
Trials started31224% of signups
Spend$8,420of $12,000 budget
Daily signups
MonTueWedThuFriSatSunMonTueWed
Send the renewal to Harbor Foods?

It sends from your Gmail, as you, with the draft attached. Nothing else leaves the project.

SENDGmail · Dana Reyes, Harbor Foods purchasing
ATTACHGoogle Drive · Harbor renewal 2026.docx
Finance check done

Ana Chen confirmed net 45 at 9:18 AM. Your rule: every customer email waits for you.

Sent · logged to the project