Model routing

The right model for every request.

Kluff reads the work, then routes it. Quick replies stay on fast models. Hard problems go to thinking models. You choose Instant, Balanced, or Thinking — and you can override any thread.

A request

Mode

Routed as

Writing · Balanced

“Write the customer a clear next step.” goes to the balanced model kept for writing work.

Instant

The fast path

Balanced

Selected for this thread

Thinking

The hard problems

From capability to everyday work

Model routing, in practice.

Model routing matches the kind of work to a model within the thread’s selected mode. Instant, Balanced, and Thinking let people choose how much reasoning a task needs without selecting an individual model for every message.

Use it for
A team with a mix of quick questions, drafting, research, coding, and deeper analysis.
Start with
Start in Balanced. Choose Instant for a quick response or Thinking when the task needs deeper reasoning.
Read the setup guide
Illustrative workflow Sample data
The request

“Compare these plans and explain their tradeoffs.”

  1. Set how much reasoning this thread needs.

  2. Classify the request as analysis, writing, coding, or another supported category.

  3. Use the model configured for that category and mode.

Analysis · Thinking

Example output
Instant
Quick responses and short lookups
Balanced
Everyday work and drafting
Thinking
Problems that need deeper reasoning

3

Modes on every thread

Instant, Balanced, and Thinking.

6

Kinds of work

Coding, writing, research, analysis, support, and general.

1

Plan for the team

Unlimited AI usage. Heavy models stay on heavy work.

How a request is routed

Named by the work, then matched to a model.

Routing looks at the request and the recent thread, not a single model for the whole company. Each kind of work has its own model inside each mode.

1

Read the request

The message, the recent thread, and any files attached to it.

2

Name the work

Coding, writing, research, analysis, support, or general.

3

Match a model

The model set for that kind of work, inside the mode on the thread.

4

Answer in private

PII is redacted first. The mode stays visible on the answer.

You stay in control

Three modes. One click to change a thread.

Leave the workspace on Balanced, then move a live call to Instant or a contract review to Thinking without changing anything else.

Instant

The fast path

Lookups, recaps, and short replies. Kluff stays on the lightest model that can finish the job.

Balanced

The default

Everyday drafting, planning, and questions. The mode a thread uses until someone changes it.

Thinking

The hard problems

Debugging, design, contracts, and analysis that should be reasoned through before anyone acts on it.

The work it knows

Routing follows the kind of task.

A coding question and a customer email should not share a model just because they arrived in the same hour.

Coding

Writing, debugging, and explaining code, queries, and scripts.

Writing

Drafts, edits, and rewrites: emails, docs, and posts.

Research

Finding and checking information, including what the team already decided.

Analysis

Numbers, documents, and options: comparisons, calculations, and findings.

Support

How something works, what a thread decided, and what to do next.

General

Everything else. Also the fallback when the kind of work is not clear.

Visible, and private

See the mode. Keep the data.

Every answer shows the mode that produced it. Admins can read usage by member and by mode, and prompts still never go to OpenAI, Anthropic, or Google.

The choice is on the answer

Instant, Balanced, or Thinking is visible on the thread, so a result is never a black box.

Private model access

PII is redacted before the request runs. Workspace data is not sent to the big model vendors.

Unlimited usage, still readable

One plan includes unlimited AI. Usage by member and by mode shows where the work actually goes.

Read how Kluff handles data

FAQ

Questions, answered.

No. Balanced is the default on every thread. Switch that thread to Instant or Thinking when the work calls for it.

It stays on the general model for the mode you chose. Kluff does not guess its way onto a specialist route.

No. Kluff uses private model access, and personally identifiable information is redacted before any model call.

No. Unlimited AI usage is included in the plan. Routing keeps the heavier models for the requests that need them.

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