AI导哈

Likes开放API MCP服务

Likes开放API MCP服务是一个MCP协议服务器，将Likes开放API封装为AI助手可调用的工具，用于查询训练活动、健康指标、训练计划、反馈和跑步能力等数据。

  • by chenwynn
  • TypeScript
  • 本地部署
  • 更新于
在 GitHub 查看 chenwynn/likes-mcp-server

使用说明

Likes Open API – MCP Server

MCP (Model Context Protocol) server that exposes the My Likes open API as tools for LLM clients (e.g. Cursor, other MCP consumers).

What's New (v1.3.0)

  • Platform admin keys: if the API key owner belongs to the admin group (qw_auth_group id 1/2), read tools can query any user_id with a date range up to 366 days (list_activities is also exempt from the 1 req/min limit). Write tools (push plans, comments, camps) keep normal coach permissions.
  • Added get_metrics tool (GET /api/open/metrics): current params, daily fitness/fatigue/performance curve, run force & threshold pace change logs, weight logs, PBs
  • list_plans: new user_id (admin only) and end params
  • list_feedback: new page / page_size params
  • get_activity_detail: documented as single-activity only — don't loop over it for long-range analysis

v1.2.4

  • Added get_period_report_json tool for GET /api/reporter/generatejson-sync
  • Added explicit permission notes for self/coached-trainee and game_id membership checks
  • Updated package description to include period report capability

Tools

Tool Description
list_activities Activity list (31-day range, 366 for admin; optional user_id for coach/admin; rate limit 1 req / 1 min, admin exempt)
get_health Health data for secondary analysis (daily hrv + sleep-related daily summary sleep). Supports optional user_id (single user) or user_ids (comma-separated batch trainees). Querying others requires coach permission for each target. Max range 31 days.
get_metrics Training parameter changes in a date range: current params, performance (CTL/ATL/TSB daily), param_changes (run_force / t_pace logs), weights, pbs. Optional user_id (coach/admin). Max 31 days, 366 for admin.
list_plans Calendar plans (from start, 42 days, or to end; 366 days max for admin). Optional user_id (admin only).
list_feedback Training feedback (start/end required, max 7 days). Optional user_ids (comma-separated) to get multiple trainees' feedback (only if you are coach of each). Each row includes plan_title, plan_content, activity (linked workout overview with system score), coach_comment.
add_feedback_comment Coach adds a comment to a trainee's training feedback. Params: content, feedback_id.
push_plans Push training plans to calendar (batch). Response includes id per result (use with delete). Optional game_id + user_ids for coach batch push to trainees.
get_game Get a training camp (game) detail and members. Requires game_id; only if you are the camp's editor or coach.
list_my_games List camps where you are editor or coach (paginated, default 10 per page, max 10).
get_running_ability Running ability: by run force (0–99) get predicted times and pace zones (E/M/T/A/I/R); or by race times (time_5km, time_10km, time_hm, time_fm, time_3km, time_mile) get estimated run force. Time format: seconds or M:SS or H:MM:SS.
get_period_report_json Generate period report JSON (/api/reporter/generatejson-sync). Requires user_id; supports range_unit, start_time, end_time, game_id. Permission: self or coached trainee; when game_id is passed, caller must be editor/coach of camp and target user must be a camp member.

Setup

cd likes-mcp-server
npm install
npm run build

Configuration

Environment variables:

  • BASE_URL – API base (e.g. https://my.likes.com.cn). Default: https://my.likes.com.cn
  • API_KEY – Your open API key (X-API-Key). Required for all tools.

Run

# After build
npm start

# Or with env
BASE_URL=https://my.likes.com.cn API_KEY=your-key node dist/index.js

Development (no build):

BASE_URL=https://my.likes.com.cn API_KEY=your-key npm run dev

Cursor integration

Add the server to Cursor MCP (e.g. Settings → MCP or .cursor/mcp.json):

{
  "mcpServers": {
    "likes-open-api": {
      "command": "node",
      "args": ["/Library/WebServer/Documents/likes-mcp-server/dist/index.js"],
      "env": {
        "BASE_URL": "https://my.likes.com.cn",
        "API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Use the absolute path to likes-mcp-server/dist/index.js. Ensure npm run build has been run first.

API reference

See the in-app docs (设置 → API 文档) or the open API routes:

  • GET /api/open/activity – activity list (optional user_id for coach)
  • GET /api/open/health – health data (hrv + sleep-related summary), optional user_id (single) or user_ids (comma-separated batch), max 31 days; querying others requires coach permission for each target
  • GET /api/open/plans – plans list
  • GET /api/open/feedback – feedback (start, end; max 7 days; optional user_ids comma-separated); response rows include coach_comment
  • POST /api/open/feedback/comment – coach comment on a feedback (body: content, feedback_id; user_id/uid from session)
  • POST /api/open/plans/push – push plans (body: plans, optional game_id, user_ids). Response results[].id is the plan id for delete.
  • DELETE /api/open/plan/:id – delete a single plan (id from push response; only if created by current key user).
  • GET /api/open/game?game_id= – camp detail and members (ownership required)
  • GET /api/open/games – list my camps (page, limit)
  • GET /api/open/ability – running ability: query by runforce (0–99) + optional celsius, or by race times (time_5km, time_10km, time_hm, time_fm, time_3km, time_mile). Returns run force, pace zones, predicted times or by-distance run force.
  • GET /api/reporter/generatejson-sync – period report JSON generation (requires X-API-Key, user_id; optional range_unit, start_time, end_time, game_id; includes coach/member permission checks).

相关 MCP