Models and cost
A model for every task, and the cost of every result
Pick the model per agent or for a single task. Every result shows what it cost, split by model tokens and machine time, so you know what you paid for.
Models and cost
The right model for each task, and what each result cost
Choose the model per agent, or for a single task. Then see the cost of every result, not just the monthly bill: which pull request, plan or dead end it paid for, split by model and machine time.
A model per agent
- agent/billingClaude Sonnet
- agent/frontendGPT
- agent/infraClaude Opus
- agent/docsGemini Flash
Just this task
agent/billing uses Gemini Flash to triage the test failures
Send routine work to a fast, inexpensive model and hard problems to a stronger one, or add your own provider with a plugin.
What each result cost, this week
$15.30 in total
- Claude Sonnet1.3M tokens$3.52
- Gemini Flash420k tokens, test triage$0.14
- microVM41 min$0.46
Give your agents badges, not keys
Kosem is in early access with a small group of engineering teams. Tell us about yours.