Any model.
Your data.
You choose who thinks for each bot, and you decide what a bot may do without asking. Everything else stays on your server.
Three sources — your pick
A model can be set for the whole team or for each bot separately. If the main one is busy, the request goes to the fallback.
On your server
Open models run on the same machine as the swarm. Nothing leaves it.
With your key
Any model that has an API. You connect it with your own key — and only if you want to.
Free ones
Online models at no cost — so you can start right away without setting anything up. The list keeps itself current: new ones appear, retired ones drop off, and today’s limits are shown next to each model.
Who decides what is allowed
Before an important step — publishing, paying, deleting — the bot sends a card. You answer once, forever, or deny. Auto-check rules set what is allowed without asking, and what never is.
Words are checked against deeds
When a bot writes “done”, VibeBot matches that against the action log. If the claim isn’t backed up, you see it in the review card right after the answer.
The reviewer
A separate model reads the original task and the finished result and says whether they match. Especially useful when the result is a contract, a report or code.
On your server
The swarm, the database, the files, the conversations and the voice all live on one server. The desktop app and the browser only show what happens there. The voice models are yours too. Backups and export are in the team settings.