How a briefing is built
The pipeline
The agents run in order: Commander → Scout → Analyst → Sentinel → Synthesizer. The Commander goes first to classify your topic; Scout, Analyst, and Sentinel each gather their own data; the Synthesizer writes the final briefing. A full run takes roughly 30–40 seconds.
Adaptive sections
The Commander classifies every topic into one of four types, and the briefing's sections adapt accordingly:
| Topic type | What the briefing emphasizes |
|---|---|
| Token | Price, liquidity, supply, holders, contract checks |
| Chain | Total value locked, native asset, ecosystem activity |
| Protocol | Protocol TVL, yields, governance |
| Narrative | Trending assets, flows, and the wider story |
Honest by construction
When a data source is slow or unavailable, the agent simply skips it and the briefing notes the gap. Confidence drops when fewer sources respond. Pyxis never invents a fallback number.
Next: Data sources & freshness →