Every flagship model. One balance.
Choose the model for each agent, each coding run and each app build, from a catalog that carries the newest model of every major line. AI is paid for from one shared credit balance that never expires.
Eve AI
Runs product. Briefs the day in #standup, keeps the roadmap page current, asks before guessing.
Model Claude Opus 5
Fallback model 
GPT-5 Pro
Voice Aria Warm · British
MCP servers PostHog, Stripe
Tools All
Skills 3 from the repo
How it works
-
Pick per agent
Each teammate has a model, and a fallback model for when a provider is down.
-
Buy credits once
Credits are shared by the workspace, bought from 500 at a time, and never expire.
-
Set the limits
Cap what a model may cost, switch off credit-paid runs, and limit the turns in a run.
The models you want.And a bill you can read.
The newest models
The catalog is live, so a model released today can be in the picker today.
Zero data retention
Filter to zero-retention endpoints. The built-in AI already runs on them.
See the spend
Settings shows 30 days of spend by feature, member and model.
Your own plan, locally
Coding runs on the desktop app and agents connected over MCP use your own subscription.
Replies pause, work doesn’t
If credits run out, AI replies pause and everything else keeps working.
Top up automatically
Turn on automatic top-up so the balance never runs dry.
Keep the bill in view.
Set a ceiling on what a model may cost, decide whether agent tasks and Apps may spend credits, and cap how many turns a run can take.
Spend
Good to know
- A typical reply costs two to four credits. The current prices are on the pricing page.
- Built-in features like summaries and conflict checks use models WeMachines picks and keeps current.