Crew
Five agents, bounded and audited. Watch one run and see exactly what each contributed.
Who does what
The Advisor has no tools at all, which is the point.
Valuer
What is it worth, and against what comparison
tools: query_sql, area_summary, hedonic_predict, index_lookup
YieldAnalyst
What does it return, and under which assumptions
tools: query_sql, yield_lookup, yield_for, dcf, mortgage
RiskAuditor
What could go wrong, and how much of it is measurable
tools: query_sql, risk_score, anomaly_lookup, forecast_lookup
Advisor
Writes the memo from what the others found
tools: none
Auditor
Checks the run against policy, whatever the outcome
tools: none
The Advisor is granted nothing — not told to avoid tools, not given them. It reasons only over what the analysts established and over the cited corpus, which is why every number in its memo is one another agent produced with a query behind it. No tool in the registry has side effects, and the Auditor blocks any run where one appears.
Run one
Running the crew needs the analyst role. Change it in the sidebar.
The run
Each agent reports as it finishes.
What it costs
Zero, by construction.