A free AI personal trainer that knows your training.

Why generic AI fitness advice falls short

Ask any general-purpose AI 'how often should I add weight on bench press?' and you'll get a decent answer — something about double progression, rep ranges, and listening to your body. The advice is technically correct. It's also completely disconnected from you.

The AI doesn't know that your bench press is already progressing faster than your rows. It doesn't know you've been training for ten months, that your volume dropped two weeks ago, or that your estimated 1RM has stalled for a month. It's pattern-matching against everything it was trained on and returning the median sensible answer. That's not coaching — it's a well-formatted Wikipedia excerpt.

A human coach charges for the listening and observing part, not the knowledge part. The knowledge is free and widely available. The scarce thing is knowing your context — your history, your weak points, your current load, your rate of progress. Without that context, even the best AI model is guessing.

What 'an AI that knows your data' actually means

The Model Context Protocol (MCP) is an open standard, published by Anthropic in 2024, that lets AI assistants read structured data from external apps through a secure, consent-based connection. It works the way OAuth works for apps: you authorize access, the AI reads what you've allowed, and you can revoke that access any time.

SammaGym provides an MCP connector at https://app.sammagym.com/api/mcp. When you connect your ChatGPT or Claude account to this endpoint, the AI gains six read tools that reach into your actual logged data. It doesn't get a username and password to your account — it gets a scoped OAuth token that lets it call specific read operations, nothing more.

Here's what those six read tools cover:

• Training summary. Sessions logged, total volume in kilograms, volume broken down by muscle group, weekly session consistency, and date of your last session.
• Active program. Your current training days, every exercise per day, sets, and rest periods — so the AI knows what program you're following, not just what you happened to do.
• Strength progress. Estimated one-rep maximums calculated via the Epley formula, best set per exercise, and bodyweight trend over time.
• A specific day's detail. Full set-by-set breakdown of any logged session — weight, reps, and RIR for every set.
• Cardio and runs. Distance, pace, speed, calorie estimates, indoor or outdoor flag, and a per-activity summary for each cardio session.
• Timed workout results. Rounds, reps, time, cap, and distance for AMRAP, EMOM, Tabata, and For Time workouts.

One important privacy note: GPS route coordinates from outdoor runs never leave your device. Only computed metrics — distance, pace, duration — are included in the data the AI can read.

What it can tell you: progress, weak points, load timing, and program critique

With access to your real logged data, an AI assistant can answer questions that would otherwise require a coach reviewing your training log manually. Some concrete examples:

• Progress analysis. 'How has my estimated bench press 1RM changed over the last eight weeks?' The AI reads your strength progress data and gives you the actual trend, not an approximation.
• Weak point identification. 'Which muscle groups am I under-training relative to the rest of my program?' It cross-references your volume per muscle group against your active program structure.
• Load timing. 'Am I ready to add weight to Romanian deadlifts this week?' It looks at your recent sets, rep ranges, RIR values, and whether you've been hitting the top of your progression target.
• Program critique. 'Is there a muscle group I'm neglecting based on what I've actually been doing?' Volume data by muscle group tells that story more accurately than a manual review.
• Recovery context. 'My training volume dropped two weeks ago — should I keep the reduced load or push back up?' Session history and consistency data let the AI respond to what actually happened, not what might have happened.

These are the kinds of questions where data access changes the quality of the answer meaningfully. The NSCA's Essentials of Strength Training and Conditioning (4th ed.) is explicit that effective programming requires monitoring both volume and intensity over time — monitoring that generic advice skips entirely because it has no data to monitor.

Research on hypertrophy training by Brad Schoenfeld consistently points to the importance of progressive overload and volume tracking as predictors of muscle growth. Neither principle can be applied by an AI — or a coach — without visibility into the actual training record.

It proposes, you approve — nothing auto-applies

If you ask the AI to suggest a new training program or modify your current one, it can do that. The write capability covers proposing a full program with rest days, warm-up sets, intensity techniques like rest-pause, cluster sets, and drop sets, pyramid rep structures, timed days, and new custom exercises with their muscle group assignments.

The critical distinction: nothing the AI proposes is applied automatically. Every proposal appears in a dedicated 'Proposals' screen inside the SammaGym app. You review the full suggested program, tap to approve each change you want, and discard whatever you don't. Until you tap approve, your active program is untouched.

This supervision layer matters. It prevents an AI mistake or a misunderstood prompt from quietly replacing a working program. You remain the decision-maker; the AI does the analytical work and drafts the proposal.

Free vs paid AI coaching apps: an honest framing

Several apps now offer some form of AI coaching, typically at a subscription cost. The range is wide — from apps that use AI to generate static programs from a questionnaire (no ongoing data access) to fully human coaching platforms that use 'AI-assisted' as a marketing qualifier.

The SammaGym approach differs in one significant way: it doesn't create a proprietary AI coaching product. Instead, it gives your existing AI assistant — ChatGPT or Claude, whichever you already use — structured access to your training data via an open standard. The AI model itself is not locked inside the app.

What this means practically:

• SammaGym is free. The workout tracker, the MCP connector, the full app — no subscription, no ads, no paywalled features.
• ChatGPT or Claude require a plan that supports custom connectors. ChatGPT's developer mode or custom connector features require a paid plan. Claude's connector integration is available on Pro, Max, and Team plans. If you don't have one of these, the connection won't work — this is a real cost and worth knowing before you commit to the setup.
• The workout log is yours regardless. You can use SammaGym as a standalone tracker — no AI connection required. The MCP connector is optional. All your data remains stored locally and in your Supabase-backed account whether or not you ever connect an AI assistant.

If you have an existing ChatGPT Plus or Claude Pro/Max subscription and already use those tools for other work, the training data connection costs nothing incremental — you're using a capability you already pay for. If you'd need to subscribe specifically to get this feature, weigh that cost against the value of AI-analyzed training advice before proceeding.

How to set it up

The connection takes a few minutes. You'll need a SammaGym account (free) and a ChatGPT or Claude account on a plan that supports custom connectors.

• 1Create a SammaGym account at app.sammagym.com if you haven't already. Log at least one workout so there's data for the AI to read.
• 2For Claude: Go to Claude.ai → Settings → Connectors → Add custom connector. Enter https://app.sammagym.com/api/mcp as the connector URL. Claude will redirect you through SammaGym's OAuth flow — sign in with your SammaGym account and authorize access. The connection completes in one click.
• 3For ChatGPT: In developer mode or via custom connector settings, add https://app.sammagym.com/api/mcp as the MCP server URL. Complete the OAuth authorization when prompted.
• 4Test the connection by asking your AI assistant something data-specific: 'What was my total training volume last week?' or 'Which muscle groups have I been training most?' If it returns real numbers from your logs, the connection is working.
• 5To disconnect: Remove the connector from your AI assistant's settings, or revoke access from within your SammaGym account. Your workout data remains in SammaGym; the AI simply loses read access.

Once connected, you interact entirely through your normal ChatGPT or Claude conversation interface. There's no separate AI coaching screen in SammaGym — the app stays a workout tracker; the AI stays your assistant. The two are connected but independent.