Pricing
What uses Draftbit credits?
Draftbit credits pay for the AI model calls that Draftbit funds, mainly agent threads that read, plan, and change your app. Each call costs the provider's token price plus a Draftbit markup, converted at 1,000 credits per USD. Threads that run on your own key or subscription do not use credits.
Reviewed September 2, 2026
Credits are spent on one thing, the AI model calls Draftbit pays for. Those happen in agent threads, and every request an agent makes while it reads files, plans, writes code, or answers a question is metered. Visual editing, previews, publishing, and the sandbox itself do not draw credits.
The charge for a call is the provider’s price for that model plus a Draftbit markup, turned into credits at 1,000 credits per USD and rounded up. Input and output tokens, including reasoning, prompt cache reads and writes, and web searches all count, so a long task that touches many files and loops through several attempts costs far more than a short question or a targeted edit. Models differ in token price. The Credits and Usage pages under Billing, and the usage details on each thread, show what a run consumed by model and provider.
Threads that run on your own OpenAI, Anthropic, or OpenRouter key, or on a ChatGPT or Claude subscription, go straight to that provider and skip Draftbit credits. That works on every plan, including Free. On Free, Draftbit credits themselves only run Codex with supported OpenAI models.
Give the agent a clear goal and the files that matter. If an approach fails twice, change the instructions rather than resending the same prompt.
Useful links
Related questions
Start with your app idea
Create a free account and describe what you want to build.