
Withings MCP
Withings in ChatGPT, Claude and every MCP app.
Withings has no official MCP server.
Memory.fit is a hosted MCP server for athletes: connect Withings in the Memory.fit app, add one address to your AI, and it reads weight, body fat and fat-free mass.
You need the Memory.fit app on Android (the iPhone app is coming), and an AI that can add a custom connector.
The address to add to your AI
https://api.memory.fit/mcp?src=mcp-withingsWithings and your AI, in short
- Official Withings MCP
- None
- Memory.fit
- Hosted MCP server, nothing to install
- Your AI reads
- Weight, Body fat, Fat-free mass
- Workouts to Withings
- No
- Works in
- ChatGPT, Claude, Gemini, Grok Bot and any MCP app
- Price
- $9 a month or $69 a year, with 7 days free
Checked October 28, 2026.
What your AI reads
What your AI can read from Withings
Once Withings is connected in the Memory.fit app, every AI you add reads the same history, under its own permissions.
- Weight
- Body fat
- Fat-free mass
- Muscle mass
- Your full weigh-in history
The tools that read it:get_daily_metricsget_trendsget_sleepget_workoutsget_source_data
Setup
Set up Withings in your AI
1. In the Memory.fit app
- 01In the Memory.fit app, open Settings → Health data.
- 02Tap Withings, then Connect.
- 03Sign in to Withings and allow access.
- 04Your whole history comes in, then every new weigh-in arrives on its own.
2. In your AI
Pick your AI. Each one signs in with your Memory.fit account.
ClaudeAny plan · about a minute
- 01Install in Claude
- 02Tap Add, then sign in to Memory.fit
- 03In a chat, turn on Memory.fit under + › Connectors
Claude adds new connectors on the web or desktop. On a phone, use a computer once, and after that it works in the Claude app too.
Every step for ClaudeChatGPTOn a computer · about two minutes
- 01On chatgpt.com, open Plugins and tap +, then Add custom MCP server
- 02Name it Memory.fit and paste the address
- 03Choose OAuth, confirm the warning, then Create as a plugin and sign in
- 04In a new chat, type @ and pick Memory.fit
The address: https://api.memory.fit/mcp?src=mcp-withings
Coming soon: one tap from ChatGPT's plugin directory.
Every step for ChatGPTGeminiGemini Spark · Google AI Pro or Ultra · US only
- 01On gemini.google.com: Settings › Connected Apps
- 02Under Custom apps, tap Add a custom app
- 03Paste the address, tap Next, then sign in to Memory.fit
- 04In a chat, type @ and pick Memory.fit
The address: https://api.memory.fit/mcp?src=mcp-withings
Grok BotComes with higher SuperGrok and Cursor plans
- 01In the Memory.fit app, open Settings → AI agents → Grok Bot and copy your Memory.fit link.
- 02Open Grok Bot on your computer or phone and message any of your Bots.
- 03Send “Add a custom MCP server called Memory.fit at” and paste your Memory.fit link.
- 04Check the name and link it shows you, then approve it.
- 05Tap Authorize and sign in with your Memory.fit account. Memory.fit works from your next message.
The address: https://api.memory.fit/mcp?src=mcp-withings
GrokSign in · about a minute
- 01Copy the address
- 02On grok.com: Connectors › New connector › Custom
- 03Paste it, then sign in to Memory.fit
The address: https://api.memory.fit/mcp?src=mcp-withings
Cursor
- 01In the Memory.fit app, open Settings → AI agents → Cursor and create your key.
- 02Add the block to ~/.cursor/mcp.json (or .cursor/mcp.json in a project).
- 03Check Cursor Settings → Tools & MCP shows Memory.fit as connected.
{
"mcpServers": {
"memory-fit": {
"url": "https://api.memory.fit/mcp?src=mcp-withings",
"headers": { "Authorization": "Bearer mem_pat_your_key" }
}
}
}Cursor connects with a key from the Memory.fit app. Paste it into Cursor's settings, never into a chat.
Every step for CursorClaude CodeOn a computer · Claude Code · about a minute
- 01In your terminal, run the command
- 02Start Claude Code and run /mcp
- 03Pick memory-fit and sign in to Memory.fit
claude mcp add --transport http memory-fit "https://api.memory.fit/mcp?src=mcp-withings"CodexOn a computer · Codex CLI · about a minute
- 01In your terminal, run the command
- 02Then run: codex mcp login memory-fit
- 03Sign in to Memory.fit in the browser that opens
codex mcp add memory-fit --url "https://api.memory.fit/mcp?src=mcp-withings"Try it
Try these prompts
“How has my weight trended over the last eight weeks?”
Uses:get_daily_metricsget_trends
“Is my body fat moving with my training volume?”
Uses:get_daily_metricsget_workouts
“Set my daily targets for race week from my current weight.”
Uses:get_daily_metricsset_day_targets
“What did my Withings scale record this month besides my weight?”
Uses:get_source_data
Your options
Every way to get Withings into an AI
Withings runs no MCP server. Community servers run on your own computer, hosted services run it for you, and so does Memory.fit.
| Option | Where it runs | What you need |
|---|---|---|
| Community serverSchimmilab/withings-mcp-serverOn your own computer | On your own computer | Your own Withings developer app, run on your own computer or in Docker. |
| Hosted servicewithings-mcp.comHosted by its author | Hosted by its author | A Withings sign-in on a server its author runs as a personal project, offered as is. |
| Hosted serviceAthleteDataHosted | Hosted | An AthleteData account. |
| Hosted MCP serverMemory.fit MCPHosted, in ChatGPT, Claude, Gemini, Grok Bot and any MCP app | Hosted, in ChatGPT, Claude, Gemini, Grok Bot and any MCP app | The Memory.fit app with Withings connected. $9 a month or $69 a year, with 7 days free. |
When another option is better
- You want only your Withings data in Claude and nothing else: the community server at withings-mcp.com connects it with a Withings sign-in.
Limits, said plainly
- Memory.fit reads Withings and sends nothing to it.
- The first weigh-in of the day counts, and readings Withings marks as a guest are left out.
- With a ScanWatch or a Sleep Analyzer, your workouts and nights come in too.
Sources, checked October 28, 2026
- github.com/Schimmilab/withings-mcp-server: Python, OAuth2 with your own Withings app; local or Docker
- github.com/akutishevsky/withings-mcp: Community hosted at withings-mcp.com/mcp; personal project, as is, no guarantees
- athletedata.health/guides/withings-mcp-server: Withings MCP Server guide
Questions
Withings MCP, in short.
No. Withings does not run an MCP server. Community servers exist, and you run them yourself.
Yes. Connect Withings in the Memory.fit app, add Memory.fit to ChatGPT as a connector, and ChatGPT reads it. Claude and every MCP app work the same way.
No. Memory.fit is hosted: you add one address to your AI and sign in with your Memory.fit account. Nothing to install, no password of another service to paste.
Each AI gets only the permissions you allow, every change shows in the app with who made it, and you can disconnect an AI or a source at any time.
Related
More MCP pages
Withings is a trademark of Withings. Memory.fit is not made by or affiliated with Withings.
Your next chapter starts with everything you've learned.
On Android now. iPhone is coming: join the waitlist and we'll email you the day Memory.fit is on the App Store.
No spam. One email when it's ready.