You pay for tokens.
That's the whole pricing page.

No. Sign up and $0.50 is added to your balance — no credit card required. How far it goes depends on the model and how large your sessions are: light work on Standard stretches much further than Pro on big contexts. Unused credit expires after 14 days.

Pure pay-per-use. Your balance is deducted after each LLM request based on tokens consumed. No subscriptions, no monthly fees. Top up any amount and use it for as long as it lasts.

Requests are paused with a clear message. No debt, no surprise charges, no work is lost. Top up to continue exactly where you left off.

After your first request in a session, the system prompt and conversation context are cached. Cached tokens cost 90% less than fresh tokens. In typical sessions, 85–98% of input tokens hit the cache — that’s why real costs are dramatically lower than raw per-token rates.

Nilux does not train any models on your prompts or your code, and does not store your code on our servers — it is held in memory for the session and cleared when the session ends. Processing itself runs on third-party AI providers under their own terms; only prompts and tool results travel over the wire, while file operations happen locally on your machine.

Yes. Standard is the default — fast and cheap for everyday work. Switch to Pro for complex reasoning tasks. Each model is billed at its own rate.

It’s based on real production data from our beta. Your costs will vary depending on session length, cache hit rate, and how many subagent tasks you delegate.