# HubFit MCP server > Connect HubFit to Claude, ChatGPT, Gemini, Grok, Muse or Perplexity with the HubFit MCP server: setup, the tools, example prompts and signing in. Source: https://hubfit.com/developers/mcp Server URL: `https://mcp.hubfit.com/mcp`. Add it as a custom connector in Claude, ChatGPT, Gemini, Grok, Muse or Perplexity and sign in with HubFit. ## Connect ### Claude 1. **Open connectors.** In Claude, go to **Settings → Connectors** and click **Add custom connector**. 2. **Add HubFit.** Name it **HubFit** and paste the server URL, then click **Add**. ```text https://mcp.hubfit.com/mcp ``` 3. **Sign in.** Click **Connect**, sign in to HubFit and approve access. Then turn HubFit on from the **+** menu in any chat. > On Claude Team and Enterprise, an owner adds the connector once under **Admin settings → Connectors**, then each person connects their own HubFit account. ### ChatGPT 1. **Turn on developer mode.** In ChatGPT, go to **Settings → Apps & Connectors → Advanced settings** and turn on **Developer mode**. 2. **Create the connector.** Back in **Apps & Connectors**, click **Create**. Name it **HubFit**, paste the server URL and choose **OAuth** for authentication. ```text https://mcp.hubfit.com/mcp ``` 3. **Sign in.** Sign in to HubFit and approve access. Pick **HubFit** from the tools menu when you start a chat. > Developer mode is available on paid ChatGPT plans. Workspace admins on Business and Enterprise may need to allow custom connectors first. ### Gemini 1. **Open connected apps.** In the Gemini web app, open **Settings → Connected Apps** and add a **custom app**. 2. **Add HubFit.** Name it **HubFit** and paste the server URL. ```text https://mcp.hubfit.com/mcp ``` 3. **Sign in.** Sign in to HubFit and approve access. Once connected, HubFit also works in the Gemini mobile app. > Custom apps in Gemini are for personal Google Accounts in the US, aged 18 or over, in English, with Keep Activity turned on. ### Grok 1. **Open connectors.** Go to **grok.com/connectors**, click **New Connector** and choose **Custom**. 2. **Add HubFit.** Paste the server URL. ```text https://mcp.hubfit.com/mcp ``` 3. **Sign in.** Sign in to HubFit and approve access. > On Grok Business and Enterprise, an admin adds the connector in the xAI console under **Grok Business → Connectors**, then each person connects their own HubFit account. ### Muse 1. **Add a custom connector.** In Muse, open **Settings → Connectors** and add a custom connector. If you don't see that option, ask Muse in a chat to add a custom connector for HubFit. 2. **Add HubFit.** Name it **HubFit** and give it the server URL. ```text https://mcp.hubfit.com/mcp ``` 3. **Sign in.** Sign in to HubFit and approve access. ### Perplexity 1. **Open connectors.** In Perplexity, go to **Settings → Connectors** and add a **custom connector**. 2. **Add HubFit.** Choose **Remote**, name it **HubFit**, paste the server URL and pick **OAuth** for authentication. ```text https://mcp.hubfit.com/mcp ``` 3. **Sign in.** Confirm the custom connector warning, click **Add**, then sign in to HubFit and approve access. > Custom connectors are on Perplexity Pro, Max and Enterprise, in the web and Mac apps. On Enterprise, an admin has to allow them first. ## Tools ### Clients | Tool | What it does | Access | | --- | --- | --- | | `hubfit_find_clients` | Search clients by name or email, list everyone with a tag, or list them all, with each full profile. | read | | `hubfit_get_client` | Get one client's profile by id. | read | | `hubfit_create_client` | Add a client, email them their login and, optionally, run an onboarding flow for them. | write | | `hubfit_list_tags` | List the tags you group clients by, with how many clients have each. | read | | `hubfit_list_onboarding_flows` | List your onboarding flows, to run one for a new client. | read | | `hubfit_list_client_summaries` | What each client did between two dates: workouts, nutrition, habits, metrics, health data and check-ins. Answers questions across clients in one call. | read | ### Check-ins | Tool | What it does | Access | | --- | --- | --- | | `hubfit_list_check_in_submissions` | List submitted check-ins, newest first: the ones awaiting review, or a client's whole history. | read | | `hubfit_get_check_in_submission` | Read one check-in in full, every question with the client's answer. | read | | `hubfit_review_check_in_submission` | Mark a check-in reviewed and send your comment to the client by email and push. | write | ### Training | Tool | What it does | Access | | --- | --- | --- | | `hubfit_list_completed_workouts` | List completed workouts with duration, totals, personal records, and the client's note and rating. | read | | `hubfit_get_completed_workout` | Read one completed workout in full, with every set the client logged. | read | | `hubfit_find_client_exercises` | List the exercises a client has logged, with their all-time records. | read | | `hubfit_list_client_exercise_history` | Every time a client logged one exercise, such as their squat over the last 8 weeks. | read | ### Nutrition | Tool | What it does | Access | | --- | --- | --- | | `hubfit_list_nutrition_logs` | List the days clients logged nutrition, with each day's totals against its target. | read | | `hubfit_get_nutrition_log` | Read one logged day: every meal, food and serving. | read | | `hubfit_get_client_nutrition_target` | Read a client's daily calorie and macro targets for each weekday. | read | ### Habits and metrics | Tool | What it does | Access | | --- | --- | --- | | `hubfit_list_client_habits` | List the habits you gave a client and each one's goal. | read | | `hubfit_list_habit_progress` | One habit's progress by day, week or month, misses included. | read | | `hubfit_list_client_metrics` | List what a client tracks, such as weight, waist or body fat, with the last entry. | read | | `hubfit_list_metric_entries` | Every entry a client logged for one metric, such as each weigh-in. | read | | `hubfit_get_client_health_data` | Daily steps, active calories, resting heart rate and sleep from Apple Health or Health Connect. | read | ### Messages | Tool | What it does | Access | | --- | --- | --- | | `hubfit_send_client_message` | Send a client a chat message from you, in your chat with them. They're notified straight away. Up to 30 a minute and 500 a day per coach. | write | ### Notes and tasks | Tool | What it does | Access | | --- | --- | --- | | `hubfit_list_client_notes` | List the notes your team keeps on a client. | read | | `hubfit_create_client_note` | Add a note to a client. A shared note shows in the client's app and sends them a push. | write | | `hubfit_list_tasks` | List the team's tasks from HubFit's Tasks page. | read | | `hubfit_create_task` | Add a to-do to your Tasks page, such as a reminder to follow up with a client. Only your team sees it. | write | | `hubfit_update_task` | Change a task's title, due date, assignee or client, or mark it done. | write | ## What you can ask ### Check-ins and accountability - "Which check-ins are waiting for my review? Summarize each one in a line." - "Summarize Maya Collins's last four check-ins and flag anything I should follow up on." - "Draft a review comment for Sam's latest check-in. Show me before you send it." ### Training - "Which clients completed fewer than two workouts last week?" - "How has Maya's back squat progressed over the last six weeks?" - "Who set a personal record this week? Give me one line on each." ### Nutrition and habits - "Who was under their protein target on most days this week?" - "Who has lost more than 5 lb this month? Give me a one-line win for each." - "Which clients missed their step goal three or more days this week?" ### Across your roster - "Who hasn't trained, logged food or checked in for 10 days?" - "Rank my clients tagged 'online' by workouts completed this month." - "Which clients slept under six hours on average last week?" ### Admin, done for you - "Add jamie@example.com as a client and run the New Online Client onboarding flow." - "Add a private note to Maya's profile: left knee niggle, swap back squats for box squats until cleared." - "Add a task for me to call everyone who hasn't checked in for two weeks, due Friday." ## Signing in Your assistant sends you to hubfit.com to sign in. You approve access and choose your workspace, and your password never touches the chat. Each team member connects their own account and only sees what their role allows. ## Security - **Your permissions, nothing more.** The assistant acts as the person who signed in. Admins reach every client; standard team members only reach the clients assigned to them. - **You approve every change.** Tools that change data are marked **write**. Assistants such as Claude and ChatGPT ask you to approve each one before it runs. Keep those approvals on. - **Only the tools you want.** Turn individual tools on or off in your assistant's connector settings. With every write tool off, it can answer questions but can't change anything. - **Watch for prompt injection.** Check-in answers, workout notes and client notes are written by people, and text in them could try to instruct your assistant. Read write actions before approving them, especially when other tools are connected in the same chat. - **Revoke any time.** Disconnect in your assistant, or under **Settings → MCP** in HubFit. - **Check the address.** Only connect to `https://mcp.hubfit.com/mcp`. HubFit's sign-in always happens on hubfit.com, never inside a chat. ## Troubleshooting - **The assistant doesn't use HubFit.** Turn the connector on for the chat (the **+** menu in Claude, the tools menu in ChatGPT), or mention HubFit in your question, e.g. "Using HubFit, …". - **Sign-in keeps failing.** Remove the connector and add it again, then sign in to HubFit when asked. - **Some clients are missing.** Standard team members only see the clients assigned to them. Ask a workspace admin to assign them, or connect with an admin account. - **"Too many requests".** HubFit allows 120 requests a minute per coach. Narrow big questions with a date range or a tag. ## FAQ ### What is MCP? The Model Context Protocol is an open standard that lets AI assistants use tools from other apps. The HubFit MCP server gives assistants like Claude and ChatGPT a safe, permissioned way to read and update your HubFit workspace. ### Which AI assistants work with HubFit? Claude, ChatGPT, Gemini, Grok, Muse and Perplexity: add HubFit as a custom connector and sign in with your HubFit account. Some limit custom connectors by plan or region: Gemini's are for personal Google Accounts in the US, and Perplexity's are on Pro, Max and Enterprise, on the web and Mac. ### Do I need to know how to code? No. In each assistant you paste one URL and sign in with your HubFit account. ### Does it work in languages other than English? Yes. Ask in any language your assistant understands, such as Spanish, French, Arabic or German. The assistant turns your request into HubFit tool calls and replies in your language, and anything it writes to a client, such as a message or a check-in review, is in whichever language you ask for. ### What can the assistant see? Exactly what you can see in HubFit, and nothing more. Admins see every client; standard team members only see the clients assigned to them. ### Can it change things without asking? Tools that change data are marked as write tools, and assistants such as Claude and ChatGPT ask you to approve each one before it runs. Only six tools change anything: adding a client, reviewing a check-in, sending a message, adding a note, and adding or changing a task. ### Can I keep it to answers only? Yes. Turn off the tools that change data in your assistant's connector settings, and it can only read. ### Does HubFit train AI models on my data? No. HubFit only sends your assistant the data a tool call asks for. How your AI provider handles that data is covered by their own terms and settings. ### How do I disconnect? Remove the connector in your assistant, or revoke it in HubFit under Settings → MCP. Access stops straight away.