The five agents
Pyxis runs five specialized agents in a fixed sequence. Each consumes a curated bundle of data and hands its findings to the next.
Commander
Commander — decomposes your topic into agent-specific queries and classifies it as a token, chain, protocol, or narrative. This classification shapes the rest of the run.
Scout
Scout — the information and narrative agent. Surfaces news, announcements, and contract-verification signals.
Analyst
Analyst — the quantitative agent. Reads price, TVL, liquidity, supply, yields, and holder concentration from structured data only.
Sentinel
Sentinel — the social agent. Reads sentiment across Twitter and Reddit, governance proposals, and a macro overlay.
Synthesizer
Synthesizer — merges the three findings into a single briefing with sections adapted to the topic type, and appends the Data Freshness table.
The pipeline is always exactly these five calls. Pyxis adds new data dimensions by expanding what each agent reads — not by adding more agents.
Next: How a briefing is built →