Find — Opportunity Radar

You don't tell the app which stock to look at. It scans your universe — daily when scheduled — using only setups with a current validation certificate for that universe (Python scores, no LLM, no cost), and surfaces what triggered and what changed since the previous scan. Everything without a certificate is Lab. Names are alphabetical inside each setup — nothing here is a ranking or a buy list. Pick a name and send it to Research.

Universe

Daily scan

Loading…

Scan options

Also show live fundamentals on cards (context only, slower)

Loading the latest scan…

Research — TradingAgents

Multi-agent research on one name: analysts gather evidence, bull and bear debate, a trader proposes, the risk team weighs in, the portfolio manager rates. Uses LLM tokens. It produces a rating and reports — it never places an order. Take a finished decision to Act.

Research setup

Ticker / Symbol

Analysis date

Analyst team

Market Social / Sentiment News Fundamentals

Deep analysis (override cost ceiling)

Loading run config…

Enter a ticker and press Run research — or start from a name on the Find page so the run carries its Radar evidence. Research started here by hand is outside the validated evidence path (the page will say so).

Act — Paper & Broker

No real money. Two paper venues: a simulated ledger (fills against real quotes) and your Interactive Brokers paper account. Position sizing, limits, stop-loss and the drawdown kill-switch are enforced in code; the server decides whether a decision's evidence allows an action. Broker orders are tracked and reconciled here.

Loading…

Review

After the fact: how stored research decisions performed, two runs side by side, LLM backtests over a date grid, and what the LLM calls cost. Read-only except LLM backtest, which re-runs the pipeline and spends tokens.

How well have the agents' stored recommendations performed? Decision-quality evaluation of past research, benchmark-relative — not a trading equity curve and not the validation that gates the Radar.

Loading evaluated decisions…

Two completed research runs side by side — ratings, decisions, verified market snapshots, reports. No LLM cost.

vs

Pick two runs to compare.

Re-runs the full multi-agent pipeline over a grid of tickers and dates, then scores the decisions. Each cell is a complete research run and uses LLM tokens. This is not the point-in-time setup validation on the Find page.

Tickers (comma-separated)

Every N days

Start date

End date

Analyst team

Market Social / Sentiment News Fundamentals

—

Read-only research record. Every portfolio strategy tested in Strategy Lab, the objective it was judged by, every stage it went through and why it passed or failed. Failed and inconclusive results are valid research outcomes, not errors. Strategies are changed only by registering a new frozen hypothesis (python -m strategylab register), never by editing here.

Loading…

Real measured LLM spend: provider token usage × the pricing registry. Unknown-model calls are shown as unpriced, never $0.

Loading spending…

Sign in to TradingAgents

Find — Opportunity Radar

Research — TradingAgents

Research setup

Act — Paper & Broker

Review

⚡ Live Agent Activity

📄 Reports

🔎 Opportunity Card

👁 Watchlist & Monitoring

📈 Forward Evidence

🕑 Run History

🔬 Strategy

🗄 Data inputs (audit)