Model Context Protocol · Hosted

The Exercise MCP Server.
For every AI agent.

Connect Claude Code, Cursor, ChatGPT and Grok to 1,900+ exercises with 720p video, a workout and program generator and supplements. Build fitness apps faster, or give your own AI coach real tools. One line to connect.

https://api.workoutxapp.com/mcp

Get a free API key →

Add to Claude Desktop Add to Cursor Add to VS Code

Claude Desktop: open the downloaded file, paste your key, done. Also: ChatGPT · Grok

Claude Code · workoutx connected

Add a workout screen to the app. Use a 45-minute dumbbell chest session from WorkoutX as the sample data.

workoutx · generate_workout{ goal: "muscle_gain", equipment: ["dumbbell"], duration: 45 }

Got a real session from the API. Main block:

Dumbbell Bench Press4 × 6–10 · rest 2 min
Dumbbell Incline Fly3 × 10–15 · rest 90 s
Push Ups3 × to near failure
Dumbbell Pullover3 × 10–15 · rest 60 s

✓ wrote WorkoutScreen.tsx · ExerciseVideo.tsx

Works with the agents developers use

Claude Code Claude Desktop Cursor ChatGPT / OpenAI Grok / xAI VS Code

What is it

An exercise database your agent can call

The WorkoutX MCP server is a hosted Model Context Protocol server at https://api.workoutxapp.com/mcp. It gives AI agents 20 read-only tools: exercise search with 720p video, alternatives, calorie estimates, a workout and multi-week program generator and supplements, plus developer tools for endpoint reference, typed models, code examples, request debugging, usage stats and database seeding.

Any MCP client that supports Streamable HTTP can connect with a WorkoutX API key. The free plan includes 500 requests with no credit card.

1,900+exercises

2,700+720p video clips

20MCP tools

7languages

01 / Connect

Set up in 30 seconds

Pick your client and paste your key in place of wx_your_key. No server to run, no SDK to install.

Terminal

claude mcp add --transport http workoutx \
  https://api.workoutxapp.com/mcp \
  --header "X-WorkoutX-Key: wx_your_key"

One command. Claude Code sends your key with every request.

claude_desktop_config.json

{
  "mcpServers": {
    "workoutx": {
      "command": "npx",
      "args": ["-y", "workoutx-mcp"],
      "env": { "WORKOUTX_API_KEY": "wx_your_key" }
    }
  }
}

One click: download workoutx.mcpb and open it. Claude Desktop installs the extension and asks for your key. Or add this to claude_desktop_config.json (uses the workoutx-mcp npm bridge).

Add to Claude Desktop

~/.cursor/mcp.json

{
  "mcpServers": {
    "workoutx": {
      "url": "https://api.workoutxapp.com/mcp",
      "headers": { "X-WorkoutX-Key": "wx_your_key" }
    }
  }
}

Click Add to Cursor, then replace wx_your_key with your key. Or paste this into mcp.json.

Add to Cursor

OpenAI Responses API

curl https://api.openai.com/v1/responses \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5",
    "input": "Build me a 45-minute dumbbell chest workout",
    "tools": [{
      "type": "mcp",
      "server_label": "workoutx",
      "server_url": "https://api.workoutxapp.com/mcp",
      "headers": { "X-WorkoutX-Key": "wx_your_key" },
      "require_approval": "never"
    }]
  }'

Use the hosted server as a tool from the OpenAI API. A one-click ChatGPT app connector is coming soon.

xAI Responses API

curl https://api.x.ai/v1/responses \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-4.7",
    "input": [{ "role": "user", "content": "What can I do instead of barbell squats at home?" }],
    "tools": [{
      "type": "mcp",
      "server_label": "workoutx",
      "server_url": "https://api.workoutxapp.com/mcp",
      "headers": { "X-WorkoutX-Key": "wx_your_key" }
    }]
  }'

Grok connects to remote MCP servers from the xAI API. xAI calls the tools for you.

.vscode/mcp.json

{
  "servers": {
    "workoutx": {
      "type": "http",
      "url": "https://api.workoutxapp.com/mcp",
      "headers": { "X-WorkoutX-Key": "wx_your_key" }
    }
  }
}

Click Install in VS Code, or add this to .vscode/mcp.json for GitHub Copilot agent mode.

Install in VS Code

STEP 1

Get a free key from the dashboard. 500 requests, no card.

STEP 2

Add the server to your client with the snippet above.

STEP 3

Ask in plain words. Your agent picks the right tools by itself.

02 / Build faster

Your coding agent knows the API

In Cursor or Claude Code, your agent calls WorkoutX while it writes your mobile or web app: real responses, typed models, working snippets per platform, and your own usage and errors when something breaks.

“Build a React Native screen that lists chest exercises with their videos.”

The agent pulls a working pattern, checks the real response, then writes the screen with the right field names.get_code_example · search_exercises

“Add Swift models for exercises and workouts to the app.”

Codable structs generated from the spec, ready to paste.generate_types

“Why are we getting 429s since yesterday?”

Your error breakdown, rate limit and quota, then the exact cause of a failing request.get_usage_stats · diagnose_request

“Seed our Postgres with every dumbbell exercise: id, name, muscles.”

CSV straight into a migration. No media links, so no media limit used.export_exercise_seed

“What should the home screen feature this month?”

The exercises and muscle groups WorkoutX apps request most, anonymised.get_trending_exercises

“Map the 40 v1 exercise ids in our database to v2 and write the migration.”

v1 → v2 matches with match quality, straight into a migration script.map_v1_exercise

03 / Ship it in your product

Give your AI coach 20 tools in one line

Building a fitness chatbot or AI coach? Add WorkoutX as an MCP tool in your OpenAI, xAI or Claude API call. Your model can search exercises, show videos and generate workouts and programs, with no integration code to write or maintain.

Same API key, plans and quotas as the REST API
Answers in 7 languages with lang
Signed 720p video links your app can play directly
Read-only tools, safe to auto-approve

server.ts · OpenAI Responses API

const res = await openai.responses.create({
  model: "gpt-5",
  input: userMessage, // "Plan my week: 3 days, dumbbells only"
  tools: [{
    type: "mcp",
    server_label: "workoutx",
    server_url: "https://api.workoutxapp.com/mcp",
    headers: { "X-WorkoutX-Key": process.env.WORKOUTX_KEY },
    require_approval: "never",
  }],
});

// res.output_text: the plan, built from real
// WorkoutX exercises with video links

04 / Tools

20 tools. All read-only.

Each tool call counts as one API request on your plan, with the same quotas and limits as the REST API. The free plan's 500 requests cover a lot of building. See pricing.

tools/list20 tools

Developer tools

get_endpoint_reference

Parameters, response shape, errors and a curl example for any endpoint.

All plans

generate_types

TypeScript, Swift, Kotlin or Dart models for every response.

All plans

get_code_example

Working snippets for React Native, Swift, Flutter, Next.js, Node and Python.

All plans

diagnose_request

Paste a failing request: get the cause and the fix. Never sends it.

All plans

get_usage_stats

Your requests, errors by status, p95 latency, top endpoints, quota and media limits.

All plans

export_exercise_seed

Ids, names, muscles and equipment as JSON or CSV to seed your database. No media. Does not use the media limit

All plans

get_trending_exercises

What WorkoutX apps request most, anonymised. Pick what to feature.

All plans

Exercise data, workouts and supplements

search_exercises

Search 1,900+ exercises by name, body part, muscle, equipment, difficulty or model. Pick fields for smaller answers. Combining filters: Starter+

All plans

get_exercise

Instructions, tips, muscles, sets and reps, calories and the video for one exercise.

All plans

get_exercise_media

Fresh signed 720p MP4 and thumbnail links.

All plans

list_exercise_filters

Every valid body part, muscle, equipment and category.

All plans

find_exercise_alternatives

Swaps for the same muscles, with or without given equipment.

All plans

find_similar_exercises

The closest movements, ranked.

All plans

estimate_calories

Calories burned from body weight and duration.

All plans

map_v1_exercise

Find the v2 exercise for an old v1 id.

All plans

generate_workout

A full session: warm-up, sets, reps, rest, cool-down and swaps.

Growth+

generate_program

Multi-week programs with periodisation and deloads.

Scale

search_supplements

Supplements with evidence ratings, dosage and timing.

Growth+

get_supplement_stack

A recommended stack for a goal.

Scale

check_api_key

Your plan and remaining quota.

All plans

05 / FAQ

Questions

What is the WorkoutX MCP server?

A hosted Model Context Protocol server at https://api.workoutxapp.com/mcp. It gives AI agents such as Claude, Cursor, ChatGPT and Grok 20 tools over the WorkoutX Exercise API: exercise search with video, alternatives, calories, a workout and program generator, and supplements.

Is it free?

Yes, you can start on the free plan: 500 requests in total, no credit card. Each tool call counts as one request. The workout generator and supplements need Growth, and programs and supplement stacks need Scale.

How do I authenticate?

Send your WorkoutX API key in the X-WorkoutX-Key header, or as Authorization: Bearer. Listing the tools works without a key, so clients can discover them first.

Why use MCP instead of calling the REST API?

Use both. The REST API is what your app calls in production. The MCP server lets AI agents use the same data: a coding agent can check real responses while it writes your integration, and an AI product can let its model call WorkoutX directly without you writing tool code.

Which clients work?

Any client that supports remote MCP over Streamable HTTP with custom headers: Claude Code, Cursor, VS Code, Windsurf, the OpenAI and xAI Responses APIs and the Claude API. Claude Desktop uses the workoutx-mcp npm bridge.

Can the agent show exercise videos?

Yes. Results include signed links to 720p MP4 animations with male and female models. The links expire, so the agent fetches fresh ones with get_exercise_media instead of storing them.

Start for free

Get a key, add one line to your client, and let your agent build with real exercise data.

Get a free API key → REST API docs

The Exercise MCP Server. For every AI agent.

An exercise database your agent can call

Set up in 30 seconds

Your coding agent knows the API

Give your AI coach 20 tools in one line

20 tools. All read-only.

Questions

Start for free

The Exercise MCP Server.
For every AI agent.