Use the workspace
Every chat, compare run, and agent call is metered as it happens.
Product · Govern
See where the spend actually goes.
AI usage analytics by model, provider, member, and project, with a monthly budget that warns before it hurts.
The invoice stopped being a mystery the first month we could point to the project behind it.
- Tokens and estimated cost per model and provider
- Breakdown by member and by project
- Monthly budget with warn and over states
100% of tokens attributedby model, provider, member, and project, because billing runs on your keys
How it works
Open usage
Spend and tokens, broken down the way you manage: by person, project, model, or provider.
Set the budget
A monthly number for the workspace. The shell shows on-budget, approaching, or over.
01
Per-seat plans hide usage. Your keys reveal it.
When billing runs on your own keys, every token is attributable. You can finally see which projects earn their spend and which model is quietly expensive.
- Cost estimates from published provider rates
- Editable pricing per model
- History by month
02
Spend less without a policy memo.
Chat suggests a cheaper capable model when one exists for the job. The suggestion shows up in the moment, not in a report three weeks later.
- In-chat cheaper-model nudges
- Value pick flagged in the model list
- Budget states visible to the owner
Questions, answered
Estimates come from published provider rates, and you can edit pricing per model when your contract differs.
The shell shows on-budget, approaching, or over for the month, and owners see the state before the invoice arrives.
Size it with a free tool
Estimate your team’s LLM API cost
Model a month of usage across every provider you have enabled.
Compare published LLM API prices
Sort input and output rates and context windows side by side.
How model-agnostic teams cut AI costs
Where the savings actually come from when you own the keys.