Research API
Base path /api/research — JWT required. Runs the research agent (Claude + domain-routed connectors) and manages saved runs. Research endpoints stream progress over SSE.
Endpoints
| Method | Path | Purpose |
|---|---|---|
| POST | /api/research | Run the research agent (SSE) |
| POST | /api/research/pipeline | Run the full research pipeline (SSE) |
| POST | /api/research/director | Run the director (multi-step orchestration, SSE) |
| GET | /api/research/saved/:projectId | List saved research runs |
| POST | /api/research/saved/:projectId | Save a research run |
| DELETE | /api/research/saved/:id | Delete a saved run |
Depth
A run takes a topic and a depth — Quick / Medium / Deep — which controls how many connectors and how much web search the agent uses. Results arrive as graded sources you can curate and bookmark.
Streaming
The POST research endpoints return Server-Sent Events: progress, per-connector results, and a final summary. Consume with an EventSource (or the platform UI).
AI usage is server-side and budget-capped — provider keys never reach the browser.
→ Related: Sources · Intelligence · Connectors