AGENTIC TRADING CO.

Command Center
Instruments scanned
Snapshot (Amsterdam)
Heartbeat
Equity
Unrealized P&L
Cost of carry
Trades · total / today
Avg conviction
Approval rate
Vetoed
Analyst avg · composite
Conviction

Conviction over time

long short no direction min to trade

Conviction histogram

Analyst avg score

−1.00 … +1.00
Guardrails · hard limits enforced in code

Account & risk

Daily drawdown vs halt threshold
0.00%

Open positions · paper account

Open positions

SymbolFilled (Amsterdam)DirectionQty EntryStop Take profitR:R MarkuPnL

Instruments · latest debate per symbol

Instrument decisions

SymbolTime (Amsterdam)Direction Conviction StageStrategy Fit Regime Analysts · T / F / S News Blackout
Why nothing traded

Veto breakdown

Strategy usage

Activity

Event feed

Recent trades

Methodology
Nine specialized agents — three analysts (technical, fundamental, sentiment), a news/blackout monitor, a strategy selector, a bull and a bear debater, a risk officer and an execution agent — run a heartbeat loop over instruments. Each pass produces a debate with a conviction score from 0 to 100; the risk officer approves only what clears the conviction floor and stays inside every guardrail on this page — risk per trade, a ATR stop, maximum leverage, and a hard trading halt at daily drawdown. The Decision Feed tab is the audit trail behind these numbers: every headline the agents read and every verdict that followed, newest first. Each entry type has its own cap there (60 headlines, 80 decisions, 30 fills, 30 exits) so a busy screener run cannot push the news out, and a veto restated unchanged on later runs is folded into one row with a ×n count. Every number on this page is fetched live from the dashboard_snapshot row in Postgres (schema_version: ), which python scripts/export_dashboard.py upserts after every pipeline run. The page re-reads that row every 60 seconds, so nothing is baked into this HTML and no state is committed to git. Fields the pipeline has not measured yet (win rate and profit factor, which need closed exits) render as , never as zero. Cost of carry is what holding costs: perp funding, fetched live per coin from Hyperliquid, and a annual borrow rate on stock shorts, accrued every mark cycle. It is booked into equity and never into a position's P&L, so realized P&L is gross and the tile shows the net. Forex carry is not modelled.
source: dashboard_snapshot (Supabase) schema_version mode
news — a headline the agents received decision — the verdict on a symbol (click to pin the detail) trade — an executed fill exit — a position closed, with its realized P&L ×n the same veto restated on later runs, folded into one row
source: dashboard_snapshot → feed[] schema_version
Agent Overview · last 3 events per agent
Agent Roles & Responsibilities
source: pipeline agent_reports + feed[] + debates[] schema_version
round 1 — opening statements by bull & bear later rounds — each side rebuts the opponent's previous argument verdict — moderator conviction after weighing the debate