Guide / Help Center

From your first New Valuation to connecting your own API — the full workflow is here.

To first understand why this workflow is designed the way it is (especially the purpose of the Locked Card and Refresh vs Revision), start with "Why You Need a Locked Card."

① Run a New Valuation

• Go to New Valuation. That page does one thing: create a card. (Updating an existing card happens on the card itself — see ③.)
• Enter a ticker and as-of date — the engine auto-fills price, fundamentals, and analyst consensus (NTM / FY EPS) from primary sources, and auto-detects the valuation model from the company’s SEC industry code (you can override it). You just set your required return; current position and your own data source (e.g., IBKR) are optional.
• Choose an output tier: Snapshot Card (fast, free trial) or Institutional Locked Card — the full 0–13 audit (subscription).
• Click "Run valuation." An Institutional Locked Card runs in the background by default — you can close the page; it completes and saves on the server.
• With "Save as my Locked Card" checked (the default), the finished card lands under My Cards.

② Store cards in your own GitHub (optional)

• By default your cards are stored in your account on the platform — nothing to configure, and you can export everything as a zip from My Cards at any time.
• Prefer full ownership in your own repository? Go to Settings → Connect GitHub storage, enter the repository (owner/repo) and a GitHub fine-grained personal access token (Contents read/write on that repo only), then connect.
• Cards you save afterward are written to your repository (path locked_assumption_cards/<TICKER>/) along with a machine-readable registry file. Disconnect at any time in Settings.
• Tip: bind a test repository first while you get comfortable, so your production card library stays clean.

③ Data Refresh (recompute on new data, keep long-term assumptions)

• In My Cards, open a ticker → "Apply new data (Data Refresh)."
• Pick the as-of date — price, fundamentals, and consensus re-pull automatically for that date; anything you typed yourself is never overwritten. Long-term locked assumptions stay unchanged.
• The refresh runs as a background job: close the page freely; the dated workpaper saves on the server.

④ Revision (assumptions changed → new version, automatic)

• During a Data Refresh the engine screens the evidence against your locked assumptions. If a locked assumption has materially changed, the same job automatically revises the valuation and saves a new version (v2/v3…) — charged once, with an Appendix A change ledger and a code-side assumption-delta check for the audit trail.
• No trigger → an ordinary refresh workpaper; your active card is never revalued without an evidence-based trigger.

⑤ Connect your own API keys (optional)

When signed in, go to Settings to enter your own keys:

• AI model: choose a provider (OpenAI / Claude / Gemini / DeepSeek / Qwen), enter your key, and optionally override the model ID. Pro members with their own key get unlimited valuations.
• Data sources: FMP (prices, consensus), Tavily (news retrieval). SEC EDGAR and FRED rate anchors are built-in official sources — no key needed.
• When the engine runs your valuation it uses your keys first; left blank, it uses the platform defaults plus free official SEC data.
• All keys are stored encrypted; the page shows only "Configured / Not configured," never echoes plaintext, and never shares them with other users.

Where to obtain each key: DeepSeek → platform.deepseek.com; FMP → financialmodelingprep.com; Tavily → app.tavily.com. Enterprise terminals such as Bloomberg/FactSet do not support simple API-key authentication at this time.

FAQ

• Why is it sometimes Provisional? When key public data cannot be fully retrieved or a mechanical QA check fails, the engine does not fabricate — it marks the output Provisional and caps the reliability score. Adding broker-grade data lifts the cap.
• The report stopped halfway? Long reports can hit a model’s single-response output limit; the engine auto-continues, but a stronger model (set in Settings) finishes in fewer rounds.
• Who owns the cards? You do. They can be stored on the platform or in your own GitHub, and exported or deleted at any time.

Still have questions? Contact us.