Option Data Import

---
name: option-data-import
description: "Validate, normalize, import, or delete an option-chain snapshot that the user is permitted to send. Use when a user attaches CSV or JSON data, asks how to format it, wants errors repaired, wants a private dashboard, or asks to delete that dashboard."
---

# Option data import

Use the hosted Option Desk MCP. Do not run a local command.

## Safety boundary

- Never fetch market data.
- Never ask for credentials, API keys, account numbers, positions, or names.
- Do not process an upload until the user confirms that they may send it.
- A user confirmation does not override the source provider's terms.
- The hosted service blocks Yahoo, yfinance, and personal Alpha Vantage data.
- Never invent a strike, quote, volatility, open-interest value, expiry, source, or rights statement.

## Workflow

1. If the format is unclear, call `option_snapshot_schema`.
2. Ask for the data source and rights confirmation if either is missing.
3. Call `option_validate_snapshot` before analysis.
4. Report each validation problem and the permitted repair. Ask for corrected data when a required value is missing.
5. Call `option_import_snapshot` only after the user also accepts storage for up to one hour.
6. Tell the user that the private dashboard URL is a bearer secret.
7. Use `option_delete_snapshot` when the user asks for early deletion.

For analysis that does not need a dashboard, use a plot or strategy tool. Those tools remove their temporary files before they respond.

## What a zero in a quote column means

Do not treat a zero bid and zero ask as an executable price of zero.
Read the validation result for the accepted price, volatility and repairs.
A row can also carry an explicit mid or user-supplied volatility.
Do not infer that the service skipped it from the bid and ask alone.

A zero bid with a positive ask differs from two zero quotes.
Report how the service handled the row without substituting your own price.

Before reporting analysis, read its degraded reason and skip counts.
State which figures are unavailable and which remain supported by accepted inputs.
Do not infer market hours or provider staleness from zero quotes alone.