AI Credits

Why a shared credit system

An AI node can call OpenAI, Anthropic, Google or a custom endpoint, in five different modes, at wildly different costs per call — a short classification is cheap, a long completion from a large model isn't. AI credits normalize all of that into one number you can actually budget against, instead of juggling four provider invoices for one workflow.

Included per plan

• • Free — 300 AI credits
• • Starter — 3,000 AI credits
• • Pro — 20,000 AI credits
• • Team — 100,000 AI credits, shared across every member

Credits reset at the start of each billing period — they don't carry over, and they're shared across every project in the workspace, not allocated per project.

How a call is priced

Cost scales with two things: the model behind the call, and the size of the request and response. As a rough sense of scale on a mid-size model:

• • Classification / extraction — cheapest; short input, short structured output.
• • Chat / completion — scales with conversation length and response size.
• • Embedding — priced per batch; embedding 1,000 short texts costs about as much as one medium chat turn.

Switching an AI node to a larger, more capable model raises the cost of every call it makes — worth doing deliberately for a step that actually needs the extra quality, not as the default everywhere.

Bring your own AI

On the Starter plan and above, an AI node can use your own API key for a provider instead of drawing from your Fimaflow credits — set it in the node's configuration, or reference one stored as an Environment Variable. Calls made with your own key are billed by the provider directly and never touch your credit balance — useful once you have your own volume pricing, or need a model Fimaflow doesn't offer through shared credits.

Tracking your usage

Remaining credits and a breakdown of recent usage — by workflow, by node — are visible from your account's billing page at any time. There is no automatic overage charge: once credits run out, AI nodes stop executing and return an error (the rest of the workflow's nodes are unaffected) until the next reset or an upgrade.