It runs on your own AI provider
SmartStrikes doesn't supply the model. AI Analysis runs on your own AI endpoint: an account with an AI provider, such as Anthropic or OpenAI, or a model running on your own computer. You enter your own API key, and the provider bills its usage to you.
This is separate from SmartAssist, the built-in help assistant, which never sees your positions. AI Analysis is the feature that reads your position data, and it sends it only to the endpoint you set up.
Setting up an endpoint
Until an endpoint is ready, the view shows AI Analysis Setup Required with a button that opens Settings → Analysis → AI Endpoint Settings. There:
- Tap Add endpoint.
- Under Start from, pick a preset: Anthropic, OpenAI, or Local (a model server on this computer, such as Ollama). The preset fills in everything but the key.
- Check the details and change the model if you like, to any model your provider offers.
- Enter your API key. Local models may not need one.
- Tap Test endpoint, and when it succeeds, Save.
An endpoint is Ready once it has its key and has passed a test. With more than one ready, Active endpoint picks the one runs use.
Endpoints are kept in your encrypted browser data. They follow you to your other browsers only if settings sync and Sync AI endpoints are both on (Settings → Security & account).
If a test fails, check the address (it usually ends in /v1), the model name and the key, and that your provider account has credit. Use the https:// address your provider gives: browsers block unencrypted http:// addresses from SmartStrikes.
Running an analysis
The AI Analysis Run Configuration card shows the endpoint in use and your accounts, each with the number of stocks that have open positions. Select one or more accounts and tap Start AI Run.
The run goes through the stocks one at a time. Run Progress shows each stock with its wheel grade and status, and fills in each review as it arrives. Cancel stops the run. When every stock is done, a summary of the whole run appears above the results. The last completed run is kept for when you come back, and Export Report saves everything as a Markdown file.
One stock at a time
Once an endpoint is ready, each stock row in the Tranches view has a sparkle button, Analyze this Position. It runs the same analysis for that one stock and shows it in a window.
What the review says
AI Analysis looks at each position the way the tranche-based wheel does, the strategy SmartStrikes is built around. Its two principles are never make a trade that loses money, and make trades that average out to your target CAGR (your Min put CAGR % and Min call CAGR % in Settings → General). In short: covered calls go on a tranche at or above its break-even, rolls are made only for a credit or even, and if the stock keeps falling, new puts go at lower strikes to open a new, lower tranche. Tranches and break-even explains those ideas.
Each stock's review has four parts:
- Position: in plain English, what's held, what each open option means, where the price sits against the strikes and break-evens, and where the position is in the wheel.
- Wheel Paths: the paths the strategy would consider now, each with its tradeoffs, such as letting an option expire or be assigned, a roll for a credit, or a new put at a lower strike.
- Covered Call Figures: for each tranche that can carry more covered calls, real call quotes from its break-even up, in the expiration closest to 14 days out, with their CAGR.
- Watch: dated facts and risks, such as an expiration, the next earnings report, a strike near the price or a wide bid/ask spread.
The run summary adds Key Risks, Time-Sensitive Items, a Monitor List and Wheel Grade Notes.
Paths, not recommendations
The paths aren't ranked and none is picked for you. The same position gets the same paths whoever holds it, because the right choice depends on your own goals and situation, which the app doesn't know. Only real quotes are used: rolls appear only when they come out at a credit or even, and calls only at or above a tranche's break-even. The model is asked to quote the app's own figures rather than work them out again.
Not investment advice. Every review is educational information written by an AI model from your position data. It isn't a recommendation to buy, sell, hold or write any security or option, and it doesn't consider your finances, goals or risk tolerance. AI can be wrong and results vary between runs, so check every figure against your broker before acting.
What data is sent
Nothing is sent until you start a run. Then, for each stock, the app sends your endpoint a text summary built from your analysis:
- a neutral name for the account, such as "Account 1", never its number, and the stock symbol;
- today's date, your CAGR targets, the stock price, implied volatility and the next earnings date, where known;
- the wheel grade and its notes;
- each open share lot and option, and the figures the app worked out: cost basis, P&L, break-evens and each tranche's details;
- option quotes the review needs: calls for tranches that can carry more, rolls that come out at a credit or even, and puts at and below the price where shares are held or an option is in the money.
The request goes from the app straight to the endpoint you configured, and what happens to it there is governed by that provider's terms.
Still stuck? In the app, tap ? to ask SmartAssist, or contact support.