FITME / MODEL CONTEXT PROTOCOL

Your training.
Connected to your AI.

Bring your own Fitme data into an AI client such as Claude. Search exercises, build plans, log workouts, and review your lifts through the Fitme MCP server.

Connect your client

An existing Fitme account and a compatible, authorized client are required.

Your AI clientSuch as Claude

Your FitmePlans & workouts

FITME MCP20 tools. One connection.

01 / CONNECT

One sign-in. Your training, in reach.

MCP stands for Model Context Protocol. Fitme’s server gives a compatible AI client tools to read and act on your training data—not just write a workout in chat. Choose your client for setup instructions.

Connect from Claude Code

Add Fitme from your terminal, then authorize it in Claude Code.

Run the command below in your terminal, outside a Claude Code session.
Open Claude Code and run /mcp. Select Fitme and follow the authentication prompt if needed.
Sign in and authorize the client on Fitme’s own consent screen, then return to Claude Code.

TERMINAL / CLAUDE CODE

claude mcp add --transport http fitme https://mcp.fitmeproject.com

Already added Fitme? Open /mcp to check its connection or sign in again instead of adding it twice.

Official Claude Code setup guide

Connect from Claude web or Desktop

Use a custom remote connector—no terminal or local server required.

Open Customize → Connectors in Claude. Choose Add custom connector (or Add → Custom → Web).
Name it Fitme and paste the server URL below.
Review the detected OAuth settings and choose Sign in now if offered. Complete Fitme’s sign-in and consent screen.
Enable Fitme from the connectors menu in your conversation before asking about your training.

REMOTE MCP SERVER URL

https://mcp.fitmeproject.com

Menu labels vary by Claude version. Team or Enterprise users may need an owner to add the connector first. Use the remote connector, not a local Desktop configuration file.

Official Claude setup guide

Connect from ChatGPT web

Add a custom MCP server where your account or workspace permits it.

Open Plugins in ChatGPT, select the plus button, then Add custom MCP server.
Name it Fitme. Under Connection, paste the server URL below and choose OAuth authentication.
Review the risk warning, select I understand and want to continue, then Create as a plugin. Complete Fitme’s sign-in when prompted.
Review the discovered tools and install the plugin. In a new chat, type @ and select Fitme.
Review and confirm any plan changes or workout writes when ChatGPT asks.

REMOTE MCP SERVER URL

https://mcp.fitmeproject.com

Availability depends on your plan and workspace permissions. Older interfaces use Settings → Apps → Create, sometimes with Developer Mode, and Scan Tools. If creation is unavailable, ask your workspace admin or check the official guide.

Official ChatGPT setup guide

Connect from Cursor

Add the remote server to your personal or project MCP configuration.

Open ~/.cursor/mcp.json for all projects, or .cursor/mcp.json in a project for project-only access. Create the file if it does not exist.
Add the configuration below. If you already have other servers, merge the fitme entry into your existing mcpServers object; do not replace them.
Save the file, then find Fitme in Cursor’s MCP settings or Customize panel. Follow its OAuth sign-in prompt and authorize Fitme.
Enable Fitme for Agent chat and review tool approval requests before it changes your training data.

MCP.JSON / CURSOR

{
  "mcpServers": {
    "fitme": {
      "url": "https://mcp.fitmeproject.com"
    }
  }
}

The configuration contains only the public server URL. Use the browser sign-in flow—do not put your password, access token, or a client secret in this file.

Official Cursor setup guide

Connect from Codex

Run the first command below to add Fitme to Codex.
If sign-in does not start automatically, run the second command. Complete Fitme’s sign-in and consent screen in your browser.
Open Codex and use /mcp to check the connection, then ask it to use Fitme for your training.

TERMINAL / CODEX

codex mcp add fitme --url https://mcp.fitmeproject.com
codex mcp login fitme

Local Codex clients, including the CLI and IDE extension, share MCP configuration on the same host. This is separate from adding a custom plugin in ChatGPT web.

Official Codex setup guide

Connect another AI client

Use a client with remote HTTP MCP and OAuth sign-in support.

Find your client’s MCP server or integrations settings and add a remote server named Fitme.
Choose HTTP (Streamable HTTP) and enter the exact server URL below.
Follow the client’s OAuth prompt. Sign in and approve access on Fitme’s own consent screen.
Check that Fitme’s tools appear. Start with a read-only request, such as looking up an exercise.

REMOTE MCP SERVER URL

https://mcp.fitmeproject.com

A client that supports only local stdio servers or static API keys cannot use this setup directly. Connection also depends on whether Clerk admits the client.

These instructions follow each client’s documented MCP setup, not a guarantee of Fitme compatibility. Your client must support remote HTTP MCP and OAuth 2.1 sign-in, and be admitted by Clerk. Client versions, account plans, and workspace policies can affect access. Need connection help?

Your account, not someone else’s. Sign-in uses Clerk and OAuth 2.1. Tools never accept a user ID, and every read and write goes through database row-level security as the signed-in user. The connected AI client receives the data its tools return; only authorize a client you trust.

02 / THE TOOLKIT

20 tools. From first plan to final set.

Browse the catalog, read your profile, manage plans, run workouts, and review exercise history. Expand an area for its exact tool names.

Catalog4 tools

Discover Fitme’s fixed choices for muscles, goals, training places, and experience.

ping: Check the connection.
list_muscle_groups: List the available muscle groups.
describe_muscle_group: Read about one muscle group.
list_plan_options: Look up valid plan goals, places, and experience levels.

Profile1 tool

Read your profile before planning or logging. Profile editing is not available through MCP.

get_profile: Read gender, age, height, weight, measurement system, and the units used for your numbers.

Exercise library2 tools

Find exercises in Fitme’s English-language library.

list_exercises: Search by name or muscle group and get exercise slugs.
get_exercise: Read the description, step-by-step markdown instructions, muscles, thumbnail, and video.

Plans6 tools

Build, manage, and switch your training plans.

list_plans: Browse your plans.
get_plan: Read a full plan, including its workout IDs.
create_plan: Save workouts, exercises, and target sets in one call. The new plan becomes active and appears in the app, with a cover photo chosen for each workout.
update_plan: Change the name, goal, frequency, workout length, or notes; optionally replace all workouts.
activate_plan: Make another plan active.
delete_plan: Delete a plan while keeping workout history.

Workout sessions6 tools

Run a workout from start to finish and browse previous sessions.

start_workout_session: Copy a workout’s exercises and target sets into a session. Starting the same workout while its session is open resumes it.
get_workout_session: Read one session in full, including its set IDs.
list_workout_sessions: Browse all, in-progress, or finished sessions with logged-set counts.
log_set: Record the reps and weight actually done for a planned set. Logging it again overwrites its values.
finish_workout_session: Close the session and move that workout to the end of the rotation.
delete_workout_session: Delete a workout session.

Progress1 tool

Look back at an individual lift over time.

get_exercise_history: Read every logged set for one exercise, newest first, to review progress or inform a suggested next weight.

03 / A TYPICAL FLOW

A conversation that becomes a plan.

“Make me a 4-day muscle-gain plan for the gym, 60 minutes a session.”

Example request—not a live conversation.

The client checks your units, finds library exercises, and saves a complete plan with target reps, rest, and weight. The plan becomes active and appears in Fitme right away.

When you train, it starts the workout, logs your actual sets, and finishes the session. That workout moves to the end of the rotation. Later, exercise history can help answer “How has my squat gone?”

FROM PROFILE TO FINISHED WORKOUT

get_profile             → units (kg or lbs)
list_exercises          → library exercise slugs
create_plan             → full plan, now active
get_plan                → workout IDs
start_workout_session   → session and set IDs
log_set (once per set)  → actual reps and weight
finish_workout_session  → finished; rotation moves on

Use returned workout and set IDs for the next calls. Use library slugs for exercises.

04 / WORKING WITH THE TOOLS

Clear rules. Fewer surprises.

Use library slugs, not exercise IDs

Refer to exercises by their library slug, such as barbell-back-squat. Plan writes validate every slug first. An unknown slug is refused by name and nothing is saved. Reads add each exercise’s name and muscle groups.

Keep the user’s units

Weights are kg for metric, or lbs for US and UK measurement systems. Heights are cm or inches. The tools do not convert values. Read get_profile first and use 0 for bodyweight.

Only one workout session can be open

Starting a different workout while a session is open is refused, with the open session named. Starting the same workout again resumes its session. Finish the open session before starting a different one.

Every write is all or nothing

Each write uses one transaction and returns its result read back. It either fully succeeds or changes nothing.

Do not assume every retry is safe

Finishing a finished session, deleting a deleted plan, activating the active plan, and logging a set again are safe repeats; repeated logging overwrites the set. Creating a plan or starting a session is not generally retry-safe: a retry can create a second one. If the outcome is unclear, inspect plans or sessions before retrying. The same-workout resume rule still applies while that session is open.

Replacing workouts creates new IDs

Passing workouts to update_plan recreates every workout, exercise, and set with new IDs. Past sessions keep their logged sets but lose their link to the old workout. Without workouts in the update, IDs stay the same. Re-read the plan after a replacement.

Read the error before trying again

Missing plans, sets, and sessions, refused starts, and unknown slugs return an error explaining the next step: which tool provides the right ID or what to finish first. If the library is unavailable, searches and plan writes ask you to try again. Reads still answer, using exercise slugs as names.

Accepted choices

Goals: muscle-gainfat-lossrecompositionstrength-performancegeneral-health-fitness
Places: gymhome
Experience: beginnerintermediateadvanced

Validation bounds

These are accepted input ranges, not training recommendations.

Plan: 1–14 workouts; 1–7 training days per week
Workout: 1–30 exercises; 1–600 minutes
Exercise: 1–20 sets
Set: 1–1,000 target reps; 0–1,000 logged reps; 0–10,000 weight in your units; 0–3,600 seconds rest
Text: Names up to 200 characters; descriptions and notes up to 2,000
Lists: 20 rows by default; up to 100 rows per page

05 / CURRENT LIMITS

Connected doesn’t mean everything.

Fitme MCP covers plan management and the planned workout logging loop. These app or Public API capabilities do not currently have an MCP tool.

It is separate from Arnold, Fitme’s AI. Your external AI client chooses how to use the tools; MCP itself is not a coaching conversation.

Edit your profile or switch your measurement system. get_profile is read-only.
Add, replace, or remove exercises, or add sets, during an open session. Only planned sets can be logged.
Preview or apply the plan update offered in the app after a finished session.
Access personal records, the weekly goal, or your streak.
Talk to Arnold, Fitme’s in-app AI.
Search or read the exercise library in Ukrainian. MCP library responses are in English.

YOUR PLAN. YOUR LIFTS. YOUR CONNECTION.

Start with Fitme.
Connect when you’re ready.

Your plan and workout history live in Fitme. MCP gives your authorized AI client a way to work with them.

Back to setup

Your training.Connected to your AI.

One sign-in. Your training, in reach.

Connect from Claude Code

Connect from Claude web or Desktop

Connect from ChatGPT web

Connect from Cursor

Connect from Codex

Connect another AI client

20 tools. From first plan to final set.

A conversation that becomes a plan.

Clear rules. Fewer surprises.

Use library slugs, not exercise IDs

Keep the user’s units

Only one workout session can be open

Every write is all or nothing

Do not assume every retry is safe

Replacing workouts creates new IDs

Read the error before trying again

Accepted choices

Validation bounds

Connected doesn’t mean everything.

Start with Fitme.Connect when you’re ready.

Your training.
Connected to your AI.

Start with Fitme.
Connect when you’re ready.