io.github.almostjacked由 GitHub 帳號 almostjacked 發布v0.3.4更新於 Oct 25, 2026
Deterministic fitness calculators — TDEE, adaptive TDEE, body fat, 1RM, macros — with consensus.
一個健身計算器伺服器，讓助理用已發表的公式計算 TDEE、體脂率、單次最大重量、巨量營養素、舉重試舉與肌肉潛力估算。
Web executable，透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式，或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南，先向你展示發布者、端點和登入方式，經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。
閱讀 https://sourceweft.com/mcp/io.github.almostjacked%2Ffitness-tools/install.md，並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前，先告訴我它的發布者、端點和登入方式，等我確認。不要把 API 金鑰寫進命令或我會提交的檔案；問我金鑰應從哪裡取得。把它新增到你客戶端的 mcpServers 設定中。
{
"mcpServers": {
"fitness-tools": {
"type": "http",
"url": "https://fitness-tools-mcp.ajwallacemusic.workers.dev/mcp"
}
}
}[CI] [License: MIT] [npm] [MCP Registry]
📖 Documentation · npm
Use it with your AI right now: add the hosted MCP connector
https://fitness-tools-mcp.ajwallacemusic.workers.dev/mcp in claude.ai / ChatGPT settings,
install one-click in Cursor
or VS Code,
or grab the Claude Desktop .mcpb.
Composable, deterministic fitness calculators — BMR/TDEE, body fat, 1RM, macros, activity multiplier, powerlifting attempts, and natural muscular potential. Each calculator runs several published formulas and reports a consensus across them.
This repo ships three separate, independently-usable artifacts:
@almostjacked/fitness-tools — the npm library. Validated,
self-describing calculators that run natively in the browser and on the server. This is
the product.@almostjacked/fitness-tools-api — a reference HTTP server built on the library.
It re-exposes the same calculators over HTTP (identical math, identical results) and is
inert until you run or host it.@almostjacked/fitness-tools-mcp — an MCP stdio server built on the library.
It exposes the same calculators as MCP tools for agents/LLM clients (identical math), runnable
with npx @almostjacked/fitness-tools-mcp.The API adds a network boundary, not capability. The package works the moment you install it; the server does nothing until it's running.
See each package's README — library · reference server.
MIT — free, open-source, and usable in commercial products. See LICENSE. Built on published formulas (Mifflin-St Jeor, US Navy, Jackson-Pollock, Epley, Casey Butt, etc.); contributions welcome.
來源：README.md，提交 ee0b492
Total Daily Energy Expenditure
Estimate BMR and TDEE via Mifflin-St Jeor, Harris-Benedict, Katch-McArdle, and Cunningham. Provide body_fat or lean_mass to unlock the LBM-based methods.
age必填sex必填height必填weight必填methodsactivity必填body_fatlean_massAdaptive TDEE
Measure actual TDEE from a logged history of daily weight and/or calorie intake. Default method (kalman) is a joint Kalman filter over true weight and TDEE — handles missing days, gates outliers, and reports uncertainty (CI95) alongside denoised true weight; regression/endpoints are simple window-based estimates kept for comparison. Use instead of formula TDEE once real logged data exists. Each entry needs at least one of weight/kcal (not necessarily both). Entry dates must span 10 years or less (earliest to latest). Note: all methods need at least one weigh-in (kalman) or two (regression/endpoints) — a history with zero weight entries throws under the kalman default; pass methods:'all' to get a graceful empty/skipped result instead.
entries必填methodswindow_daysprior_tdee_kcalBody Fat Percentage
Estimate body-fat % via US Navy circumference, Jackson-Pollock 3-site skinfold, and Deurenberg (BMI-based) methods.
agehipsex必填neckwaistheightweightmethodsskinfold_sumOne-Rep Max
Estimate 1RM from a submaximal set via Epley, Brzycki, Lombardi, Wathan, O'Conner, and Mayhew; returns a %1RM load chart.
reps必填weight必填methodsMacronutrient Split
Compute protein/fat/carb grams for a calorie target using the g-per-kg-bodyweight method, with goal-based protein defaults.
goalweight必填methodscalories必填fat_g_per_kgprotein_g_per_kgActivity Multiplier
Estimate the TDEE activity multiplier via the classic lookup table or a NEAT+EAT model (occupation/steps for non-exercise, training volume for exercise).
bmrweightmethodsintensityoccupationsteps_per_dayactivity_levelsession_minutessessions_per_weekPowerlifting Meet Attempts
Deterministic opener/second/third attempts from an estimated 1RM, plus a warmup ramp and per-side plate loading. Tunable by aggressiveness and available plates.
bar_weightone_rep_max必填aggressivenessavailable_platesMaximum Muscular Potential
Estimate drug-free maximum bodyweight at a target body-fat % via Casey Butt (wrist+ankle), the FFMI~25 natural cap, and Berkhan's max-contest-weight model. Men only in v1.
sex必填anklewristheight必填methodstarget_body_fat_pctFat-Free Mass Index
Compute the Fat-Free Mass Index (FFMI) and its height-adjusted form from weight and body fat (or lean mass), and flag whether it exceeds the ~25 natural ceiling.
height必填weight必填methodsbody_fatlean_massRelative Skeletal Muscle Index
Estimate the appendicular skeletal muscle index (RSMI) from a DXA value (direct) and/or an anthropometric estimate (Wen 2011), and flag low muscle mass against EWGSOP2, AWGS, and Baumgartner sarcopenia cutoffs.
age必填sex必填asm_kgheight必填weight必填methods已滿足適用檢查項的 100%
該評分衡量此收錄條目的完整程度，依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價，也不表示伺服器安全或目前可用：可連通性是帶日期的檢查結果，評分也不包含安裝量或使用者評價。
packages/core @almostjacked/fitness-tools — the library (the product)apps/api @almostjacked/fitness-tools-api — a reference HTTP server over the libraryapps/mcp @almostjacked/fitness-tools-mcp — an MCP stdio server over the librarycorepack enable # pnpm version is pinned in package.jsonpnpm installpnpm -r test # core + api + mcp test suitespnpm -C packages/core build # build the librarypnpm -C apps/api dev # run the reference server on :8080