OpenSysLab has one price; your AI provider has another. Understanding both keeps your monthly bill boring, which is how bills should be.
The server: fixed and predictable
You pay the monthly price of the spec you chose — from $6 for the smallest up to the top-tier options, shown on the product page and again in My Servers. There are no per-seat charges, so adding your tenth colleague changes nothing. Billing runs while the server exists, so deleting it is how you stop.
The AI tokens: metered by your provider
The Vibe-code agent talks to the model vendor using your API key. That vendor bills you for tokens — a long configuration session costs more than a quick question, and different vendors price very differently. Two things keep this cheap in practice:
- Coding-oriented plans like Kimi for Coding are priced for long agent sessions — that is why it is the default.
- The agent runs on your server, so routine operations (restarts, imports, log checks) cost fractions of a cent, not subscription dollars.
Setting a ceiling
Every major provider lets you cap usage or set billing alerts in their dashboard. Do this on day one: it converts a runaway loop in an agent script from a shock into a stopped tool.
A realistic month
A small team running a CRM with daily agent assistance typically spends more on coffee than on tokens. The server line — $24 a month for a solid 2 vCPU / 4 GB machine — is the number to budget around.
Where to see it
Server cost: the product page and your My Servers total. Token cost: your provider’s dashboard, which is the only place that number exists — we never mark it up or meter it.
