xtsebas/chatbot-server

mcp-personal-trainer

A Model Context Protocol (MCP) server over STDIO that exposes personal trainer tools:

• compute_metrics — BMI and BMR (Mifflin–St Jeor)
• recommend_exercises — recommendations by goal/sport
• build_routine_tool — weekly routine (sets/reps/timing)
• recommend_by_metrics_tool — recommendations using metrics + goal

The server does not speak HTTP; an MCP host invokes it via STDIO.

Requirements

• Python 3.10+
• pip and the venv module
• (Optional) API keys if your planner.py uses them (e.g., API Ninjas)

Installation

# 1) Clone the repo
git clone <your-repo>.git
cd <your-repo>

# 2) Create a venv
python3 -m venv .venv
source .venv/bin/activate        # Windows PowerShell: .\.venv\Scripts\Activate.ps1

# 3) Install dependencies
pip install -U pip
# If you have a requirements.txt in the repo:
pip install -r requirements.txt
# If not, install the basics:
pip install "mcp[cli]>=1.2.0" pydantic httpx python-dotenv

Expected minimal files:

server.py                     # this MCP server (calls mcp.run())
planner.py                    # implements compute_bmi, build_routine, etc.

Run in dev mode (without a host)

Use the MCP CLI to start the server in dev mode:

mcp dev server.py

This launches the process over STDIO and shows helpful logs. Stop with Ctrl+C.

Connect to any MCP host

Configure your preferred MCP host to run a STDIO server with:

• Command: python server.py stdio
• Working directory: the repo folder
• Environment: optional (TRAINER_TOOL_TIMEOUT, external keys, etc.)

Each host has its own UI to register MCP servers. Just ensure it executes that command in the repo directory.

MCP JSON examples

All messages use LSP-like framing:

Content-Length: <N>\r\n
\r\n
<JSON>

1) Initialize

{"jsonrpc":"2.0","id":1,"method":"initialize",
 "params":{"capabilities":{},"clientInfo":{"name":"my-host","version":"1.0"}}}

2) List tools

{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}

3) Call each tool

compute_metrics

{"jsonrpc":"2.0","id":3,"method":"tools/call",
 "params":{"name":"compute_metrics",
           "arguments":{"sexo":"male","edad":28,"altura_cm":175,"peso_kg":78}}}

recommend_exercises

{"jsonrpc":"2.0","id":4,"method":"tools/call",
 "params":{"name":"recommend_exercises",
           "arguments":{"objetivo":"gain muscle mass","deporte":"calisthenics","limite":10}}}

build_routine_tool

{"jsonrpc":"2.0","id":5,"method":"tools/call",
 "params":{"name":"build_routine_tool",
           "arguments":{"objetivo":"volleyball","dias_por_semana":4,"minutos_por_sesion":60,"experiencia":"intermedio"}}}

recommend_by_metrics_tool

{"jsonrpc":"2.0","id":6,"method":"tools/call",
 "params":{"name":"recommend_by_metrics_tool",
           "arguments":{"sexo":"male","edad":30,"altura_cm":170,"peso_kg":90,"objetivo":"fat loss","limite":8}}}

The server responds with {"jsonrpc":"2.0","id":<id>,"result":...} or {"error":...}.

Try a MCP remote

connect to a finance MCP with this url

https://finance-mcp.finance-xtsebas74.workers.dev/regla_503020

And try the tool for the 50-30-20 rule

Troubleshooting

• The process seems unresponsive
  Make sure you send messages with the correct framing (Content-Length + body) and read exactly that many bytes.

• Timeouts
  Increase TRAINER_TOOL_TIMEOUT (per tool). If your host has its own timeout, tune it there too.

• Missing keys/datasets
  If planner.py uses external APIs/datasets, define the required env vars or paths.

• Some tool doesn’t appear in tools/list
  Ensure @mcp.tool() decorates it and server.py ends with mcp.run().