- The answer, as the model gave it, in markdown.
- The entities in it (brands, companies, products), with their position, mention count, sentiment and stance.
- The sources the model cited.
- The fan-outs: the web searches the model ran while it answered.
- The products it showed, with prices, ratings and merchants.
- The ads shown with the answer.
How it works
1
Create a task
POST /v1/tasks with a prompt, a model, a country and a language. The task is charged $0.013 and queued.2
Wait for it
Poll
GET /v1/tasks/{taskId}, or pass a webhookUrl and receive task.completed.3
Read the result
The result stays readable for 7 days after completion. A failed task is refunded.
Models
Every model costs $0.013 a task, all enrichments included.GET /v1/models gives the live list, with prices and the countries each model runs in.
Next steps
Quickstart
Run your first task in five minutes, free, with a test key.
Task results
The answer, entity, source, product and ad schemas.
Webhooks
Get results pushed to you. No polling.
API reference
Every endpoint, with a playground.