What TrakAI actually does
Here is the whole thing in one sentence: TrakAI takes everything Reef Trak knows about your tank and packs it into a document you can hand to any AI. Your last 90 days of parameters with how they have been trending. Your dosing schedule, including the products and amounts. Every fish, coral, and invert you own, with scientific names and how long you have had them. A month of maintenance logs with timestamps and quantities. Feeding records. If you snap a photo of your tank, that goes in too. Then it wraps the whole thing in a persona prompt and your specific question.
You paste that document into whichever AI you trust. ChatGPT, Claude, Gemini, Perplexity, a local LLM running on your own machine, or whatever frontier model ships next week. The AI has the full picture: parameters, trends, dosing, livestock, maintenance, photos, all of it, in one shot.
That is the entire workflow. No subscription. No in-app chatbot to maintain. No vendor in the middle.
Why bring-your-own-AI beats closed in-app chatbots
Pop open any other reef app with AI and it is the same shape. A chat window inside the app, the vendor picks the model, the prompt is locked, you pay a monthly fee on top of whatever you already paid for the app. That works fine for surface questions. The cracks show the moment you ask anything serious.
Line it up dimension by dimension and the picture is pretty clear:
| Dimension | Closed in-app AI | Reef Trak TrakAI |
|---|---|---|
| LLM choice | Whatever the vendor wired up, usually older or cheaper to control cost | User picks any frontier model: GPT-5, Claude Sonnet 4.7, Gemini, local LLMs |
| Prompt quality | Vendor template, frozen | User crafts per-question, iterates in conversation |
| Data scope | Whatever the vendor pipeline exposes | Full 90-day structured export with livestock ages, measurement stats, maintenance quantities, feeding logs |
| Recurring cost | Subscription that compounds over the life of the tank | Zero. User already pays for their AI service |
| Future-proofing | App must ship an update to use a new model | Auto-upgrades. User moves to the next best model the day it ships |
| Audit and transparency | Black box. User cannot see the prompt or the data sent | User sees the exact markdown document being sent |
| Multi-engine cross-checking | Single vendor opinion | Paste the same export into ChatGPT and Claude, compare the answers |
Reefers tend to distrust black boxes and want to own their own data. TrakAI was built for that instinct.
Try it on a real problem
Say your alk has been bouncing. You open Reef Trak, tap TrakAI Export, snap a quick photo of your tank if you want, and copy. Paste it into Claude or ChatGPT with something like, "my alk has been bouncing the last two weeks, what should I check?"
Now the AI is working with 90 days of alk readings, your dosing schedule with product names, your livestock list, what you added recently, your water-change cadence, and the photo. It can connect dots a stock chatbot cannot. Something like: you added a hammer coral 18 days ago, bumped your dose by 0.4 ml/day 14 days ago, and your water changes dropped from weekly to biweekly around the same time. That correlation is the whole answer, and it is only possible if the AI can actually see your tank.
You get an answer grounded in your reef, not a recycled reef-keeping pamphlet.
Which AI engines work with TrakAI
TrakAI exports as plain markdown with optional photo attachments. Which means any AI that takes text and images works. That is basically all of them right now:
- ChatGPT. Paste it in. Attach the photo if you have one.
- Claude. Same drill. Solid pick for the longer-context analysis reef chemistry usually turns into.
- Gemini. Same drill.
- Perplexity. Handy when you want answers that link back to real reef-keeping sources.
- Local LLMs like Ollama or LM Studio. Paste into your local chat. Nothing leaves your machine.
- Whatever AI ships next year. The export is just markdown, so anything that reads text will handle it.
Is bring-your-own-AI a step backward?
The marketing pitch for in-app chatbots is convenience. Tap a button, get an answer, done. Fair enough for casual stuff.
But once your reef gets serious, the closed model starts to break. You cannot follow up across sessions because the chat history is gone. You cannot ask a second AI to double-check the first. You cannot use the model that just dropped this week. And you have no idea what data was actually sent or what prompt the vendor used.
TrakAI gives you the data, hands you the prompt, and steps out of your way. Yes, you do one extra paste. For anyone running a real reef with real money in the rocks, that paste is worth doing once a week to get an answer that is actually useful.
How to actually use it
TrakAI is already in Reef Trak on iPhone, Android, and the web. Open a tank, find Tank Tools, tap TrakAI Export. Reef Trak builds the document, drops it on your clipboard, and you paste it wherever you want.
No setup. No account to link. Nothing to subscribe to. It is just text on your clipboard.