What the AI Agent does

The AI Agent runs in the background, crawling websites you configure, analyzing content for relevance to your team's interests, and automatically adding the best finds to your sessions - complete with tags.

Step 1: Add your LLM key

1. Go to Settings → AI Agent
2. Add an API key from one of the supported providers:
   • OpenAI (GPT-4o recommended)
   • Anthropic (Claude)
   • Google (Gemini)
3. Your key is encrypted and stored securely

Note: The AI Agent uses your own LLM key for analysis. Curyloop never stores or shares your key in plain text.

Step 2: Configure sites to crawl

Add websites the agent should monitor:

• Tech blogs: techcrunch.com, thenewstack.io
• Reddit: reddit.com/r/programming
• Industry sites: arxiv.org, hacker-news.firebaseio.com
• Company blogs: engineering.example.com

The agent supports both RSS feeds and HTML scraping.

Step 3: Define your topics

Tell the agent what your team cares about. Topics are natural language descriptions:

• "AI and machine learning developments"
• "React and Next.js best practices"
• "DevOps and infrastructure automation"
• "Security vulnerabilities and patches"

The more specific your topics, the better the agent's relevance scoring.

Step 4: Run or schedule

• Manual run: Click "Run Now" to trigger an immediate crawl
• Scheduled: The agent runs automatically on a schedule

Each run crawls configured sites, scores content against your topics, and adds high-relevance items to your current active session.

Monitoring

• Check the agent dashboard for run history and statistics
• Items added by the agent are tagged automatically
• If the agent adds too much noise, refine your topics or raise the relevance threshold

Tips

• Start with 3–5 high-quality sources and expand gradually
• Review agent-added items weekly to calibrate relevance
• The agent learns from your topics - be specific about what matters