Skip to content

SET MCP reference.

SET MCP reference: Streamable HTTP endpoint, OAuth and bearer-code authentication, permissions, ten tools, and plan-creation behavior.

Connect your AI

Connection

Endpoint: https://fitness.olaurent.com/api/mcp

Transport: remote MCP over Streamable HTTP. Authentication: OAuth, or a SET connection code in Authorization: Bearer <code>. Configure the endpoint in a client that supports this transport and one of these authentication methods.

Use AI connections to copy installation instructions, create a named code, inspect last use, or revoke access. Follow the setup guide for the dashboard steps.

Permissions

  • training:read: read saved history, Board exercises, catalog matches, splits, and plans.
  • body:read: read profile and body data, including recorded scan metrics.
  • training:write: create exercises and training plans.

OAuth requests are limited by the approved scopes. Connection codes grant all three scopes and remain usable until revoked. Treat a code as an account credential.

Revoking the connection stops future authenticated access through it. Previously returned data remains subject to the receiving assistant’s storage and privacy settings. Revocation does not delete exercises or plans already created.

Ten tools

SET MCP tools and required permissions
ToolScopeUse
get_training_summarytraining:readSaved training totals, latest activity, and active split.
list_workout_setstraining:readSaved sets, newest first, with exercise and workout context.
list_exercisestraining:readExercises already on your Board, including stable IDs.
get_exercise_historytraining:readSaved sets for a Board exercise ID.
search_exercise_catalogtraining:readCanonical catalog matches for a search query.
list_splitstraining:readTraining splits, scheduled days, and assigned exercises.
list_training_planstraining:readCurrent and past plans, including their origin.
get_body_databody:readProfile, weight history, questionnaire, Board muscle scores, and physique scan metrics.
create_exercisetraining:writeAdd a private Board exercise after catalog search and user agreement.
create_training_plantraining:writeCreate a private plan and split; optionally replace the active plan and split.

Reading history correctly

list_workout_sets and get_exercise_history accept a limit of 1–200, defaulting to 50, and an exclusive before timestamp in Unix milliseconds. Results are newest first. Exercise history requires an ID from list_exercises.

search_exercise_catalog accepts a required query and a limit of 1–50, defaulting to 20. It searches the catalog, not the user’s history.

Body data preserves unknown values as null. Board muscle scores come from logged sets and body stats; they are distinct from physique scan appearance metrics. The body-data tool returns metrics, not photo files.

Creating a plan

Search the catalog before choosing exercise names. Ask the user to approve the complete plan before calling create_training_plan. Required fields are title, weeks, and days. A plan supports 1–52 weeks and 1–14 days; each day has a name, weekdays, and exercises with name, muscle group, load type, and set count.

activate defaults to true. Set it explicitly to false to create a proposed plan without replacing the active plan and split. Activation supersedes the existing active plan. The resulting plan and split are private.

User approval is an instruction to the assistant, not a separate confirmation dialog enforced by the tool. Grant write access only to a client you trust to follow your instructions.

Plan creation is not an update operation. Do not retry blindly after an uncertain response: first inspect existing plans to avoid creating duplicates. The MCP tool set has no delete or undo operation.

Current boundaries

MCP reads saved account data. It cannot see local unsaved sessions, log live sets, control the phone screen, operate timers, or edit an existing plan. Mobile access and remote agent control depend on the external client.

For practical prompts, see plan from your history and use AI at the gym. Return to SET for AI.