Documentation / Agents
Agents
Three agents run on a schedule and do your research automatically: the X Curation agent polls your watched accounts, the Research agent reads newsletters and portals, and the Recommendation agent builds your monthly buy/sell plan. You monitor and steer them from one console.
Before you start
- Enable at least a few X accounts on X Accounts before running the X Curation agent.
- Lookback controls how many hours back an agent searches for content. Use a large window (48–168 h) on a first run or after a long gap; keep it short (12–24 h) for daily runs to avoid reprocessing items you have already seen.
- The Recommendation agent works best after the Knowledge Base and Feed have accumulated some signal (run X Curation and Research first).
How to use it
- Go to Agents to see the status of each agent (last run, schedule, pending proposals, success rate).
- Click "Run now" on any agent to trigger it immediately — no need to wait for the schedule.
- Watch the live log stream while the agent runs.
- Check Inbox for proposals, and Feed for curated items.
- Use the prompt editor on each agent card to adjust its behaviour without touching code.
What comes next
Tips
- Run agents in dry-run mode first when changing a prompt — you see what they would produce without any changes being applied.
- The global kill switch pauses all agents immediately if something unexpected happens.
This page describes a paid feature
Everything documented here is included in the single €19/month plan.