API Reference

All REST endpoints exposed by the TI Agent backend.

All endpoints run on http://localhost:3001.

Method	Path	Description
GET	/health	Server health check, returns DB path
POST	/api/fetch-feeds	Fetch and parse RSS/Atom feeds for a date window
POST	/api/analyze	Proxy to Anthropic /v1/messages (API key auth)
POST	/api/analyze-cli	Proxy to local Claude Code CLI (subscription auth) — same response shape as /api/analyze
GET	/api/config	Load all saved config
PUT	/api/config	Save config (API key always excluded)
GET	/api/config/export	Download full config + custom feeds as JSON
POST	/api/config/import	Restore config from exported JSON
GET	/api/feeds/custom	List custom feeds
POST	/api/feeds/custom	Add a custom feed
PUT	/api/feeds/custom/:id	Rename or re-point an existing custom feed
DELETE	/api/feeds/custom/:id	Remove a custom feed
GET	/api/reports	List all reports (no body field)
GET	/api/reports/:id	Get a single report with body
POST	/api/reports	Save a generated report
DELETE	/api/reports/:id	Delete a report (cascades to its hypotheses)
GET	/api/costs	Get cost ledger rows
GET	/api/reports/:id/export?format=	Export report (md / html / pdf / docx)
POST	/api/reports/:id/hypotheses	Replace a report’s stored hypotheses
GET	/api/reports/:id/hypotheses	List a report’s stored hypotheses
GET	/api/hypotheses/counts	{ [report_id]: count } for every report in one query
GET	/api/hypotheses/consolidated?reportIds=&theme=	Consolidated hunting-hypotheses HTML (theme = light | dark, default light)

POST /api/fetch-feeds

Request body:

{
  "feeds": [
    {
      "id": "talos",
      "name": "Cisco Talos",
      "xmlUrl": "https://blog.talosintelligence.com/feed",
      "url": "https://blog.talosintelligence.com"
    }
  ],
  "dateFrom": "2025-01-01",
  "dateTo": "2025-01-07"
}

Response:

{
  "articles": [
    {
      "title": "Article title",
      "url": "https://...",
      "date": "2025-01-03",
      "summary": "First 800 chars of article text...",
      "feedName": "Cisco Talos",
      "feedIndex": 1
    }
  ],
  "errors": [
    { "feed": "Some Feed", "error": "Timeout" }
  ]
}

Feed fetcher behaviour:

• All feeds fetched in parallel with Promise.allSettled
• 8-second timeout per feed via AbortController
• Capped at 15 articles per feed
• Articles outside the date window are filtered out
• Failed feeds appear in errors[] and are noted in the report