Quick Start

Run TI Agent and generate your first briefing.

Running

Open two terminal windows in the project directory:

# Terminal 1 — backend (port 3001)
node server.js

Expected output:

📂  DB path: /Users/you/ti-agent/ti-agent.db
📦  DB file size: 32768 bytes
📦  Database ready at /Users/you/ti-agent/ti-agent.db

✅  TI Agent server → http://localhost:3001

# Terminal 2 — frontend (port 3000)
npm start

Then open http://localhost:3000 in your browser.

First Briefing

1. Go to the ⚙ CONFIG tab
2. Enter your Anthropic API key — kept in memory only, never written to disk
3. Select your feeds and date window
4. Click ⚡ GENERATE BRIEFING

The button cycles through two phases:

• ◌ FETCHING N FEEDS… — downloading and parsing RSS/Atom feeds
• ◌ ANALYZING WITH AI… — Claude processing the articles

How It Works

Browser (React)
  └─ POST /api/fetch-feeds
       └─ Express :3001
            1. Fetch all selected feeds in parallel (8s timeout)
            2. Parse RSS 2.0 / Atom 1.0 XML
            3. Filter articles to date window
            4. Cap at 15 articles per feed
            └─ Return { articles[], errors[] }

  └─ POST /api/analyze
       └─ Express :3001
            Proxy to Anthropic API with prompt:
            - Numbered article list [1]..[N] with title/URL/date/summary
            - Citation rule: cite [N] after every claim
            - Structured report sections
            └─ Claude response streamed back

  └─ Browser renders report
       - rebuildReferences() strips Claude's References section
         and rebuilds it from the verified source map
       - renderMarkdown() renders [N] as clickable links
       - Report saved to SQLite with source_map JSON