CoNudge + your AI
The structured training memory any brain can use — your AI thinks along, CoNudge validates and remembers.
Your AI thinks along. CoNudge remembers.
You probably already talk to an AI assistant — Claude, ChatGPT, or another one. It knows your context, your preferences, the way you ask questions. But a chat conversation is not a training program. It validates nothing, it logs nothing, and next month it has forgotten what you lifted last month.
CoNudge is the other half: the structured training memory. Every program, every session, every measurement and every evaluation lives in one place — validated against real training principles and your own history. The thinking can happen anywhere. Recording, validating and remembering happens in CoNudge.
The assistant is interchangeable. The memory is not.
Three ways to use it
It is a spectrum, not a menu. You decide, moment by moment, how much your own AI does.
CoNudge on its own
You use CoNudge as-is: intake, program generation, training, logging, an evaluation after every block. No external AI needed — CoNudge always works standalone.
Collaboration — the second opinion
You let your own AI look at your program or progress. A fresh perspective from the brain that already knows you. The result comes back into CoNudge as a proposal: you see exactly what would change and confirm or reject it.
Your AI thinks, CoNudge records
You live in your own assistant and use CoNudge as the system of record. Progress questions, program ideas, planning — all there. But every program lands validated in CoNudge, and you log your training in the app.
How it works today: copy and paste
The second opinion works today, with any assistant — Claude, ChatGPT, or whatever you use. No integration required:
Via the “Second opinion” card on your Train or Progress page you export a pseudonymized summary: your program, your progress, your goal — without your name or contact details.
Ask your question: “take a look at this program”, “where is the weak spot?”, “build an alternative for week 3”.
CoNudge reads the proposal, validates it against real training principles and your history, and shows you a diff.
Nothing changes until you approve. Never before.
The connector: no copy-paste (in development)
We are building a direct connection (via MCP, an open standard) that lets your assistant read along live: your current program, your profile, your progress. Questions like “how am I doing?” get answered by your AI straight from your CoNudge memory — and it can submit a program idea as a proposal itself.
To be honest: the connector is not live yet. Once it ships, you connect it in your assistant’s settings and it plays by exactly the same rules as the paste route: reading and proposing is allowed, saving directly is technically impossible.
Until then, the copy route above works with every assistant — today.
Privacy is the design, not the fine print
Your AI thinks along without you giving your data away. Three hard rules, built into the architecture:
How we handle your data is fully covered in our privacy policy.
For personal trainers: your method, your clients, your AI as assistant
As a personal trainer or online coach you may already use your own AI for your thinking. CoNudge makes that safe and useful: per client you export a mandatorily pseudonymized briefing — training data, no identity — and spar with your own assistant about the next block.
You bring the result back as a proposal. CoNudge validates it, you approve it, and only then does it go to your client. Your client management, program history and evaluation loop stay in CoNudge — your AI becomes your assistant, not your replacement.
New to CoNudge? Read why trainers build their training programs here.
Why not just do everything in the chat?
Because building a good training program is more than getting a good answer. A chatbot gives you a workout; a workout builder with memory gives you progression. CoNudge validates every program (duplicate exercises, equipment you don’t have, weights that don’t match your history), builds in week-by-week progression and deloads, computes your e1RM trends, and evaluates every block on six criteria.
You get the best of both: the free-form thinking of your own AI, and the structure, validation and progress tracking of a real training system.