Install the MCP server

Connect skcal to Claude, Claude Code, or ChatGPT over a remote MCP server.

skcal exposes a remote Model Context Protocol server so AI clients can log and query your data. It uses Streamable HTTP and is guarded by OAuth 2.1: you sign in once in the browser to authorize, with no token to copy or paste.

Claude Code

claude mcp add --transport http skcal https://app.skcal.fit/mcp

Then run any skcal request in Claude Code; the first call opens a browser to authorize.

Claude (desktop & claude.ai)

Open connector settings

Open Settings → Connectors and click Add custom connector.

Paste the URL

Enter https://app.skcal.fit/mcp and save.

Authorize

Connect and sign in with your skcal account when prompted.

ChatGPT

Enable Developer mode

Open Settings → Connectors → Advanced, and turn on Developer mode. (Requires a paid plan.)

Add the server

Click Add custom connector and enter https://app.skcal.fit/mcp.

Authorize

Available tools

21 tools, all scoped to your account.

Food and body

Tool	What it does
`skcal_get_status`	Calories + protein logged vs. targets for a day, plus latest weight
`skcal_get_targets`	Daily calorie/protein targets and goal/start weight
`skcal_list_weigh_ins`	Recent weigh-ins (pounds)
`skcal_log_weight`	Record a weigh-in
`skcal_log_measurement`	Record a body measurement
`skcal_log_meal`	Log a food item with calories + protein
`skcal_log_workout`	Log a workout from a plain-text description
`skcal_list_meals`	List meals logged on a day
`skcal_move_food`	Move a meal or item to another day

Goal phases

Tool	What it does
`skcal_get_phases`	Bulk, cut, and maintenance phases with planned vs. actual weekly rate, plus estimated maintenance calories
`skcal_start_phase`	Start a phase; closes the open one and updates your targets
`skcal_update_phase`	Edit a phase in place
`skcal_end_phase`	End the open phase

See Goal phases for how rates and calories are set.

Training

Tool	What it does
`skcal_create_mesocycle`	Plan a mesocycle from a split template or your own days
`skcal_get_training_session`	A session’s exercises, targets, logged sets, and last week’s numbers, or a whole-mesocycle overview
`skcal_list_exercises`	The exercise library, built-in and custom
`skcal_log_sets`	Log lifted sets for an exercise in the current session
`skcal_set_muscle_feedback`	Record soreness, pump, joint pain, or workload for a muscle
`skcal_complete_session`	Finish a session, log it as a workout, and plan next week’s same day
`skcal_swap_exercise`	Replace an exercise for the rest of the mesocycle
`skcal_end_mesocycle`	End the active mesocycle early

See Training for how each week is planned.

How authorization works

The server advertises OAuth discovery at /.well-known/oauth-authorization-server and /.well-known/oauth-protected-resource. Clients register dynamically, run the PKCE authorization-code flow against your skcal login, and receive a token scoped to your account. It’s the same login the apps and CLI use.