Plan

Know where your AI budget went.
Before finance asks.

Spend targets let you set a monthly dollar budget per group, track actual vs plan in real time, and get alerted when something is drifting — before the month ends and the damage is done.

Start free →View quickstart

Per-group targets

Real-time projections

Watch alerts before you exceed

The problem

Observability tells you
what happened. Planning tells you what should.

Knowing where your AI spend went is the start. Deciding where it should go — and catching drift before month end — is the harder problem.

Problem 01

AI spend has no budget until it's over

Most teams track AI costs reactively — the bill arrives and you explain it. There's no moment where you decide "search should cost $400 this month" and then hold the line. The number just grows until someone notices.

Problem 02

Month-end surprises are avoidable

By day 9 you're already on pace to overspend — but nothing flags it. A projection at current burn rate would have told you on day 3. Without it, you find out on day 31 when the invoice is already paid.

Problem 03

You can't explain AI allocation to finance

When finance asks why the AI line item went up 60% this quarter, you need more than a total. You need to show which group drove it, whether it was planned, and what you're doing about it.

How it works

Set it once.
Track it all month.

No config files. No deploys. Click the target column on any group and type a number. The dashboard immediately starts tracking actual vs plan and projecting month-end spend at current burn rate.

Tag calls with a group

Pass any string as the group — a feature name, a team, a workflow. Every call with that tag counts toward its target.

Set a monthly dollar target

Click the target in the dashboard and type a number. $400 for search. $200 for summarize. Whatever makes sense for your product.

Watch the projection update in real time

As calls come in, the projected month-end spend updates automatically. Watch catches drift before you exceed. Over target fires when you already have.

Spend targets · May 2026day 9 of 31

GroupMTDTargetProjectedStatus

search$186$400$641Watch

summarize$103$350$355On track

chat$59$400$203On track

embeddings$14$100$48On track

search is projecting $641 against a $400 target — at $20.70/day it will exceed its budget around May 19. Still 10 days to act.

Status states

Three states. One is actionable before it's too late.

Watch is the state that matters most — you're not over budget yet, but you will be. That's the window to act.

✓

On track

Actual MTD spend is under target and projected month-end spend is within 5% of target.

Nothing to do. The group is running as planned.

Watch

Actual MTD spend is under target but projected month-end spend exceeds target by more than 5%.

You have time to act — investigate the group, reduce call volume, or adjust the target. This is the state where intervention matters.

Over target

Actual MTD spend has already exceeded the target.

The budget for this group is blown for the month. Use the data to explain why and set a more accurate target for next month.

How projections work

Simple math.
Useful early.

The projection is straightforward: divide MTD spend by days elapsed, multiply by days in the month. No ML, no forecasting model — just the current burn rate extended to month end.

That simplicity is the point. On day 9, you have 9 data points. The projection tells you whether your current trajectory is heading toward trouble — with enough time left to change it.

Example · search · May 9

MTD spend$186.30

Days elapsed9 of 31

Daily avg$20.70 / day

Projected$641.70

Target$400.00

Exceeds on~May 19

Updates on every new call

As events come in, the daily average recalculates and the projection updates. If usage spikes today, you see it today.

Resets with the calendar month

Targets and MTD spend reset on the 1st of each month. Projections start fresh from day 1 of each new cycle.

"Exceeds on" date

For any Watch or Over target group, the dashboard calculates the specific date when spend will hit the target at current burn rate.

Unplanned spend section

Groups generating spend with no target set appear below the main table. Add a target to move them up. Useful for catching new features or groups you forgot to plan for.

Spend targets vs enforcement — what's the difference?

Spend targets

For your team — internal planning

Sets expectations, not hard limits

Alerts when drifting, never blocks

Per group — search, summarize, chat

Visible in the dashboard, not the SDK

Enforcement

For your users — customer-facing limits

Sets hard caps that block or warn

SDK-level — fires before the API call

Per user or per tier — free, pro, enterprise

Configured in dashboard, enforced in SDK

Most teams use both. Spend targets for internal accountability — "is search on budget this month?" Enforcement for customer-facing limits — "can a free user spend more than $5?" They solve different problems at different layers. Learn about enforcement →

Know where your AI budget is going.
Before the month ends.

Set your first spend target in 60 seconds. No config files. No deploys.

Start free →Read the docs

Know where your AI budget went.Before finance asks.

Observability tells youwhat happened. Planning tells you what should.

Set it once.Track it all month.