Research API reference
The Research API is designed for multi-source work that can take longer than one synchronous generation request.
Resource model
The Alpha v1 architecture includes a research resource/run model for creating a research job, reading state, and retrieving results or source metadata when enabled.
Confirm exact paths and schemas in the current Developer Portal/OpenAPI.
Create a research job
A request can include concepts such as:
- research question
- depth
- source/query limits
- project or organization context
- output/report preferences
The create response should identify the durable research job rather than blocking until every search and verification step is complete.
Lifecycle
A research job can include stages such as:
Sources
Where the API exposes sources, preserve source URL/title/date/metadata needed by your application to display provenance.
A source being retrieved does not itself prove that every generated claim is supported by that source.
Results
The final result can be a report or structured research output depending on the active contract.
Limits
Research can be constrained by source count, query count, runtime, model/tool availability, credits, and organization policy.
Webhooks
Use completion/failure webhooks for backend automation rather than polling indefinitely when the event catalog supports research events.
Applications making important decisions should inspect source quality and citation support rather than treating a COMPLETED research run as proof that every claim is true.