Do not build the data pipeline.
Fifteen years of options contracts, with every reading already computed on top of them. Reach it over HTTP against a published specification, or hand your assistant the MCP server and let it do the asking.
The API and MCP server are on the Vega plan
$ curl https://optionomics.ai/api/v1/overview \-H "X-USER-EMAIL: [email protected]" \-H "X-USER-TOKEN: $OPTIONOMICS_API_KEY"{"session": "open","regime": "risk_on","net_premium": 418204119,"put_call_ratio": 0.82}
The boring part is done.
The expensive part of options research was never the model. It is the eighteen months of ingesting, cleaning, partitioning and backfilling that come before you get to write one.
- The archive already exists
- Every contract, expiry, price and open-interest figure since 2011 — and nothing has ever been deleted from it.
- The derivations already ran
- The computed readings are stored per stock rather than left for you to rebuild from raw trades on every query.
- The trades are already scored
- Unusual activity, the very large trades and the flow aggregates all arrive already measured against each stock's own baseline.
- The backtests do not cheat
- History is cut off strictly before the day being scored, and entries only happen at prices that were genuinely available — the leakage bug you would otherwise have written yourself.
Your agent reads the market.
Point your assistant at the server and it does the querying itself — no client to write, no schema to keep in sync.
- Is anything unusual happening in NVDA this week? Reads the trades we have already scored, not the raw list
- Has this happened before, and what happened next? Goes to the public record and comes back with it
- Are traders betting up or down on Apple today? Where the money went, and when it turned
- Did anyone in Congress trade anything on my watchlist? Insider and congressional filings, matched to the stock
- What moved the chip stocks since the open? The whole sector, summarised in a paragraph
- Market data, and nothing else
- It answers questions about the market, never about an account. That boundary is policy rather than convention, and it is enforced in one place for every client.
- A stable surface
- Clients discover it once and cache what they find, so nothing reshuffles underneath a running session.
- Every call logged
- The tool, the arguments and the response, in your own console — including the difference between a refusal and a failure.
Webhooks with receipts.
You do not have to keep asking. Alerts post to your endpoint, and every attempt is recorded along with the response your server gave back.
Start on Vega.
Vega opens the REST API, the MCP server and strategy testing on top of everything else. You issue and revoke your own keys from your own console.
Cancel anytime · Month to month · Switch tiers whenever