Vynflow's AI API turns a single HTTP call into reliable, typed JSON. You define the shape of the answer, send a prompt and your data, and get back exactly the fields you asked for — no parsing, no prompt wrangling, no model guesswork.
POST to /api/ai/complete with a prompt, an optional set of named data inputs, and a list of response_fields. Vynflow returns a strict JSON object that matches the fields you requested — and nothing else.
POST /api/ai/complete
{
"prompt": "Classify the sentiment of this review.",
"response_fields": [
{"name": "sentiment", "type": "string"}
],
"data": []
}
// response
{
"sentiment": "positive"
}
Add as many fields as you need — strings, numbers, booleans — and pass named data items so the model has the context it needs. The response always conforms to the schema you sent, so you can deserialize it straight into your own types.
Every request is analyzed, routed, and decomposed automatically. You write the prompt; Vynflow handles the rest.
Each request is classified by complexity and routed to the cheapest model tier that can handle it — and the savings of lighter models are passed straight on to you.
Complex requests are automatically broken into up to five focused subtasks, processed independently, then synthesized into one coherent result.
Test schemas, prompts and data at /app/ai with zero code — iterate on your response fields and see results instantly before you ship.
Every request logs token usage, cost, the routing decision — complexity, tier and reasoning — and exactly which models were used.
Responses are guaranteed to match your response_fields schema, so AI slots cleanly into typed code instead of fragile string parsing.
The AI API is built for service-to-service use. Authenticate with a domain-scoped service-account API key, and every call is metered, logged and billed against your plan.
AI calls are billed per token, with the discount of cheaper model tiers passed on to you.
On the Single Free Tier, AI requests draw from your prepaid credits.
Paid plans include free token allowances each cycle, with per-token billing beyond them.
One endpoint, your schema, the cheapest model that gets it right. Start free and call the API today.