Workout MCP: Plan, Log, and Analyze Training from Claude or ChatGPT

Workout MCP connects your real training data — plan, history, saved templates, PRs — to Claude or ChatGPT, so you can plan a training week around your recovery, log a session the moment you finish it, and dig into your exercise history without opening a separate app. This guide covers Cora's actual training tools, real example prompts, and an honest look at where AI-in-chat genuinely beats a workout app's UI and where it doesn't. Current status: live in production — see the full picture in our MCP overview.

Most workout apps are built around one assumption: you're standing in a gym, phone in hand, tapping through a pre-built session. That's still the right interface for the middle of a set. It's a much worse interface for the two moments that actually determine whether your training works over months — deciding what this week should look like, and understanding what the last few weeks actually did to your body.

That's the gap workout MCP closes. Cora's training data — your plan, your history, your recovery trend — becomes something you can reason about in a normal conversation with Claude or ChatGPT, with the AI able to write changes straight back into your actual plan.

What Cora's training MCP tools actually expose

Cora's MCP server includes nine dedicated training and workout tools, scoped under training:read and training:write permissions:

Get your training plan. Your current plan and scheduled workouts, including the workout IDs needed to look up detail on any one of them.
Get your workouts. Completed sessions — both ones you followed from a plan and ones you logged manually — over a chosen date range.
Get workout details. Full exercise, set, and rep breakdown for one specific past workout.
Get saved workouts. Your saved workout templates, filterable by name.
Get cardio goals. Your weekly cardio zone-minute targets, if you've set any.
Log a workout. Takes a description of a completed strength or resistance session and logs it as a structured entry.
Manage your training plan. Add, move, rename, or remove individual workouts within your current plan.
Generate a training plan. Builds a full personalized plan from a freeform description of your goals; it replaces your current plan (the old one is archived, not deleted) and takes roughly 30 to 60 seconds to generate.
Add a workout from a template. Schedules one of your saved templates onto a specific date.

The split between "manage your training plan" (targeted edits) and "generate a training plan" (a full rebuild) matters in practice — one is for nudging your existing week, the other is for starting over. Knowing which one you're asking for helps you get the response you actually want.

Planning a training week from chat, around recovery

This is the workflow where connecting training data to AI pays off the most, because a good training week depends on synthesizing several things at once: what you did recently, how recovered you are right now, and what you're trying to accomplish. Cora's training tools, combined with its sleep and recovery tools (with the right scopes granted), let an AI do that synthesis in one pass.

Real prompts mapped to real tools:

"My HRV has been low for three days and I slept badly last night. Rework this week so today and tomorrow are easy, and push my hard session to Thursday." → reads recovery/sleep data, then calls the plan-management tool to move workouts.
"Build me a new 4-day plan: two lower body, one upper body, one conditioning day. I want to peak strength in about six weeks." → calls the plan-generation tool with your stated goal and timeline.
"What does my saved 'push day' template actually include? Schedule it for Friday." → pulls the saved template's details, then schedules it onto that date.
"Add an easy 30-minute Zone 2 session Wednesday without touching anything else this week." → a targeted addition through plan management, not a full regeneration.

Live now — in production

Everything described here is real and live in production today, at https://api.purplepill.ai/api/mcp/. Add it as a custom connector in Claude or ChatGPT and sign in with your existing Cora account. It isn't listed in any AI connector directory yet, so add it manually for now — full setup at corahealth.app/mcp.

Logging workouts by chat, right after you finish

The immediate practical win is speed. Right after a session, describing what you did in plain language is often faster than opening an app and rebuilding the workout tap by tap.

Example logging prompts:

"Log today's workout: back squat 4x6 at 225, Romanian deadlift 3x10 at 155, leg press 3x12, walking lunges 3x20." → parsed and saved as a structured logged workout.
"I did a 5-mile run at an easy pace, felt good the whole way." → logged as a cardio session with your described effort.
"Actually, change yesterday's bench press to 4 sets instead of 3 — I forgot one." → an edit to the already-logged entry.

רוצה שCora תעזור עם זה?

נסה את Cora בחינם

This works because the logging tool is doing exactly what the app's own workout logger does — parsing a structured session and saving it — just from a text description instead of a form. If you already have a training habit built around progressive overload, chat-based logging is a fast way to keep your history accurate without breaking your post-workout routine to fill out an app screen.

Exercise history and PR analysis

Where a workout app shows you a single exercise's chart if you dig for it, an AI connected via MCP can synthesize across exercises and time periods in a way that would take real manual effort in an app UI.

Prompts for history and PR analysis:

"Look at every workout with a back squat in the last four months. Am I actually progressing, or have I plateaued?"
"What was my best deadlift session this year, and what did the rest of that training week look like?"
"Compare my total training volume this month to last month."

The AI pulls the relevant workouts, reads full set-and-rep detail from each one, and reasons about the trend directly in the conversation — you can immediately follow up ("what should I do differently next block?") without re-running a separate analysis. If you're building toward a specific lift goal, our one-rep max calculator is a useful companion for setting the target the AI can reason against.

Template building and reuse

Saved workout templates are one of the more underused features in most training apps, mostly because building one manually takes effort. Through chat, template creation becomes a conversation:

"Look at my saved templates — do I already have a lower-body template, or do I need to build one?"
"Schedule my 'full body strength' template for Monday, Wednesday, and Friday next week."

Where AI-in-chat beats an app UI, and where it doesn't

This is the honest core of the comparison, and it cuts both ways:

Chat wins for:

Synthesizing recovery, sleep, and training history into one planning decision, instead of you mentally cross-referencing three separate screens.
Fast, freeform logging right after a session, especially for people who find tapping through a workout builder tedious.
Multi-week or multi-exercise analysis you can immediately question and refine in the same conversation.
Rebuilding or adjusting a plan based on a stated constraint ("I'm traveling next week and only have dumbbells") without you manually re-planning every day.

The app UI still wins for:

In-the-moment execution: rest timers, quick weight/rep entry between sets, and seeing your training plan laid out visually for the day.
Visual exercise demonstrations and form cues, which a chat window can't show you mid-set.
Anything that benefits from glanceable, at-a-gym speed rather than typed or dictated conversation.

The practical pattern that works best: plan and analyze in chat, execute in the app. That's exactly how Cora's AI coaching is designed to complement, not replace, the in-app training experience — see our broader guide on how AI fitness coaching works for more on that split.

Setting it up

Setup is the same OAuth flow used for every Cora MCP connection — add the live server URL (https://api.purplepill.ai/api/mcp/) in Claude or ChatGPT, sign in with your existing Cora account, and choose your consent bundles. "Read my data" (checked by default) covers viewing your plan, history, and templates; "Log & edit my data" is required for the AI to log workouts or change your plan. Full setup steps are in our health MCP guide and on the Cora MCP page.

רוצה שCora תעזור עם זה?

נסה את Cora בחינם

Key Takeaways

Workout MCP connects Claude or ChatGPT to your real training plan, history, and templates through nine dedicated tools — reading your plan and past sessions, logging new workouts, and rebuilding or adjusting your week.
Planning a training week from chat works best when it's grounded in your actual recovery and sleep trend, not just a stated goal in isolation.
Logging by chat is often faster than an app's workout builder for typical sessions, though very granular set-by-set RPE tracking still favors the in-app logger.
Chat-based AI is strongest at synthesis and analysis; the app UI is still better for in-the-moment execution — rest timers, set entry, and visual form cues.
This is live in production today, with scoped, revocable consent for reading and writing your training data — connect it by adding the endpoint to Claude or ChatGPT as a custom connector.

Frequently Asked Questions

What is a workout MCP, and how does it differ from a workout tracking app?

A workout MCP (training MCP) connects your training data directly to an AI assistant like Claude or ChatGPT through the Model Context Protocol, so you can plan, log, and analyze workouts by talking, not tapping. Cora's exercise MCP tools let a connected AI see your current training plan, your full workout history, and saved templates, and — with your permission — log completed sessions or restructure your plan, all inside a normal chat conversation instead of a dedicated app screen.

Can AI actually build a good training week from a chat conversation?

Yes, when it has real data to work from. Cora's training MCP tools let an AI see your recent workout history, your recovery and sleep trends, and your existing plan before generating anything, so a request like "rebuild my week around the fact that I slept badly the last two nights" produces a plan grounded in your actual recent performance and readiness, not a generic template. The plan gets written directly into your Cora app, not just described in the chat.

Is logging a workout by chat as accurate as logging it in the app?

For most people, describing a session in plain language ("upper body push, bench 3x8 at 185, overhead press 3x10, dips to failure") captures the same information you'd tap into a workout builder, often faster. The AI parses your description into a structured logged workout using the same logging tool the app itself would call. Where it's less precise is very granular set-by-set RPE tracking or complex supersets — those are still easier to enter directly in the app's workout logger.

Can an AI analyze my exercise history and PRs through MCP?

Yes. Cora's training MCP tools include a dedicated call for full workout details (every exercise, set, and rep from a specific session) plus your completed workout history over any date range, so an AI can be asked to trace how a specific lift has progressed, flag when you hit a new best, or compare your current training block to a previous one — all from your actual logged numbers.

Where does AI-in-chat training beat a workout app, and where does it fall short?

AI in chat is stronger at synthesis across data types — combining your recovery, sleep, and training history into one reasoned recommendation — and at fast, freeform logging. It falls short where a workout app's UI is genuinely better suited to the task: real-time set-and-rest timers during a session, visual exercise demonstrations, and quick taps for adding weight to a set mid-workout. Use chat for planning and analysis before and after training, and the app itself for the actual set-by-set execution.

Is Cora's workout MCP available to everyone right now?

Yes. It's live in production, not a preview — any existing, onboarded Cora account can connect it today by adding the endpoint to Claude or ChatGPT as a custom connector. There's no button inside the Cora app for this yet, and it isn't yet listed in any AI connector directory, so you add the URL yourself. Full setup steps are at corahealth.app/mcp.