Measurement

Your provider sends a total. Spaces sends you the breakdown.

One monthly invoice tells you nothing about where the money went. Spaces captures cost at the token level and tags every event with full context — task, workflow step, model, user, plan — so you can roll it up along any axis and run operations on real data, not estimates.

Project planCost
Search Overhaul v2
$770
Auth Service v2
$580
Payment Integration
$712
Agent Rollout Pilot
$525
Checkout Redesign
$570
Demand Forecasting v2
$395
PlanningImplementationReviewTesting
Total spend$3,5526 plans · 3 models
By step
Planning
$460
Implementation
$1,800
Review
$695
Testing
$597
By model
Claude Sonnet 4.6
$1,980
Gemini 2.5 Pro
$980
GPT-4o
$592

A monthly total is a symptom, not an answer. Cost only becomes manageable once it carries the context of the work that created it.

What the invoice says

One number per month, per provider. No task, no step, no owner. By the time it arrives the work is weeks gone, and you reconstruct attribution from memory and guesswork — if at all.

What Spaces records

Every usage event captured at the source, the instant it happens — input tokens, output tokens, model, and cost, each one tagged to the task, step, user, and plan that triggered it. The breakdown is already there before the bill is.

How it works

Cost intelligence, built into the workflow

No SDKs to wrap around API calls, no billing logs to parse. Spaces reads token usage straight from your coding sessions and tags every cost with the full context of the work that triggered it.

Capture

Every coding session reports token usage at the source — input, output, model, cost. Nothing reconstructed from invoices.

live session
Claude Sonnet 4.6
18,402 in2,930 out
$0.42
Tag

The same reading carries its full context — which task, step, person, plan, and model. Cost per unit of work, not per workspace.

$0.42 · one event
planPayment IntegrationstepImplementationuserPriya IyermodelClaude Sonnet 4.6
Roll up

Thousands of tagged events aggregate along any axis into one reconciled total — by plan, step, user, or model.

$0.42$1.18$0.07$0.93$0.31
$3,552all events · this window
Any dimension

Roll it up by anything you track.

Every record carries its full context — the space, the plan, the task, the workflow step, the agent, the model, the person. Ask what your AI cost any way you want — the answer’s already in the data.

By Plan
$3,552
total spend · 8 plans

Roll up every task in a plan to see what it actually cost — not what you estimated.

Top 6 of 8 by cost
Search Overhaul v2
$770
Payment Integration
$712
Auth Service v2
$580
Checkout Redesign
$570
Agent Rollout Pilot
$525
Demand Forecasting v2
$395
Then go cross-dimensional

Pick one axis, then segment it by another. Cost by workflow, then by step within each one. Cost by plan, then by model. When a plan is sourced from a document, the trail runs all the way back — a real measure of how well your documentation set up the work. The combinations surface what any single flat summary hides.

Control & optimization

Beyond visibility into control

Tracking cost is the foundation. Rolling it up along the right axes — and acting on what you find — is where the value compounds.

Budget guardrails

Set spending limits at any level — task, plan, user, or model. Alerts fire as a budget approaches its limit, not after it’s exceeded.

Cross-plan comparison

Compare actual cost profiles across plans to spot efficiency patterns and outliers — which plans run expensive, and which deliver the most output per dollar.

Spend forecasting

Real cost-per-task-type profiles feed the projection engine, so you forecast remaining spend on in-progress plans from data, not estimates.

Operational insight

Combine dimensions to surface patterns flat summaries hide — cost by step within a plan, by model within a user — and find where the real optimization lives.

Stop guessing what agents cost. Start seeing the full picture.

Every token attributed to the task that consumed it. Every cost rolled up by plan, user, workflow, and model. Every budget tracked before it’s breached — cost intelligence built into the work, not bolted on after.