Everything the terminal does, and how it decides what to show you.

A single dense workspace instead of eight browser tabs. Windows tile on a free-form desktop, linked in context groups so clicking a symbol in one moves the others with it. Layouts save. The mouse is optional.

Six rules govern the software, and they are constraints on us rather than promises to you:

1. Look around — already done for you. A new account opens on a populated workspace built from the bundled data snapshot, not on an empty grid and a setup wizard. A persistent badge tells you it is snapshot data and gives its build date; that badge cannot be dismissed.
2. Run one command. Press Cmd K (or Ctrl K), type NVDA, press Enter. You now have a chart. Type NVDA FA and you have the financial statements next to it.
3. Connect a key, when you want data newer than the snapshot. The key screen shows what each provider unlocks and what it costs at that provider, validates the key with a real round trip, and shows you the actual response. A key that fails its probe is not accepted. Adding a model key also lifts the copilot cap on every tier, including free.

Steps one and two need no account at all once the public demo is open, and no step needs a card.

The command bar is the front door. Focus it with Cmd K, type, press Enter. Four grammars are accepted, so muscle memory from a legacy terminal transfers directly.

Every window below exists in the product. Where a window is a simulation or depends on a connected key, the scope column says so rather than leaving you to find out.

Every analytic in the product has a written method with its formula, its assumptions, its failure modes and its literature reference. Twenty-five are documented. The full text ships with the product; this is the index of what is covered.

• Cross-sectional factor model, in the style of a commercial risk model
• Cross-sectional factor returns
• Risk decomposition into factor and idiosyncratic components
• Brinson-Fachler allocation and selection attribution
• Ledoit-Wolf shrinkage covariance
• Exponentially weighted covariance
• Principal-component factor covariance

• Probabilistic Sharpe ratio
• Deflated Sharpe ratio
• Combinatorial purged cross-validation
• Information-coefficient quintile spread
• Alpha decay
• Block-bootstrap Sharpe confidence interval
• Drift monitoring

• Conditional value-at-risk, also called expected shortfall
• Cornish-Fisher adjustment to value-at-risk
• Magdon-Ismail expected maximum drawdown
• Stress windows and crisis replay

• Square-root market impact
• Capacity estimation
• Kelly criterion
• Volatility targeting

• Hierarchical risk parity
• Black-Litterman
• Mean-variance optimisation

Everything the interface does, it does over a documented HTTP interface, so anything you can see you can also fetch. API access is a paid entitlement: read-only on the founding tier at 60 requests a minute, read and write on PRO at 300. Tokens are per organisation, scoped, revocable and never a session cookie.

The surface, by area:

• Market data — quotes, bars, chains, curves, reference
• Company — description, statements, estimates, holders, filings, peers
• Runs and metrics — list, series, sparkline, tearsheet artifacts
• Backtest jobs — submit, stream progress, cancel
• Analytics engine — every lab, each as its own endpoint
• Portfolio and paper books — positions, marks, staged tickets
• Alerts and monitoring — rules, fired alerts, acknowledgement
• Scheduled jobs — definitions, state, history
• Copilot — sessions, turns, streamed answers
• Outbound webhooks — per-event routing for your own endpoints
• Configuration — preferences, limits, entitlement map with live usage
• Streaming — websocket channels for quotes and job progress

Errors are a single shape with a machine-readable code, a human-readable reason, and, for a quota error, which counter was hit, what the limit is and when it resets. There is no endpoint that submits an order anywhere.

This index is maintained by hand from the documentation that ships in the repository: the window catalog, the command reference, the methodology text and the API reference. A generated command page — built from the alias map itself so it cannot drift — is planned and will replace the command section above.

If something here does not match what the product does, the product is right and this page is a bug. Send it to [email protected] and it gets fixed with a changelog line.