Life Hub
БесплатноНе проверенEnables deployment of a remote MCP server on Cloudflare Workers without authentication, allowing connection to AI Playground or Claude Desktop.
Описание
Enables deployment of a remote MCP server on Cloudflare Workers without authentication, allowing connection to AI Playground or Claude Desktop.
README
A personal remote MCP server (Cloudflare Worker) that lets Claude log and read structured life-tracking data — starting with nutrition — into an Obsidian vault.
You type a freeform meal in Claude ("3 eggs, oatmeal, a chicken breast, rice"); the model parses it and estimates macros, then calls a tool; the Worker writes one immutable event file to Cloudflare R2; Obsidian (via the Remotely Save plugin) pulls those files into a vault folder, where a Dataview dashboard summarizes them.
Nutrition is the first tracker; the design generalizes to future entry types (journal, workout, etc.) via a shared envelope + a typed body.
How it works
Claude app ──parse+estimate──▶ MCP tool ──▶ Worker ──write──▶ R2 (one file per meal)
│
Remotely Save ◀──sync┘
│
▼
Obsidian vault + Dataview dashboard
Storage model
- One immutable event file per submission. Never a read-modify-write of a shared daily note.
- Daily/weekly totals are derived at read time (and in the dashboard), never stored.
- R2 key mirrors the vault subpath, keyed by logical date so "last 7 days" is a cheap prefix-list:
Trackers/Nutrition/entries/2026-06-23/2026-06-23-190712.md - Filename stem
{logical_date}-{HHMMSS}is the idempotency key. A ULID lives inside the file as a stable internalid.
Entry format
Each entry is markdown: a YAML frontmatter envelope (common to all types) + a typed body + a human-readable line.
---
id: 01J8XQ2K7E9...
type: nutrition-entry
date: 2026-06-23 # logical date (6am-Toronto rule)
logged_at: 2026-06-23T23:07:12Z
source: claude
meal: dinner
calories: 820 # derived from items
protein_g: 63
confidence: medium
input_text: "3 eggs, oatmeal, a chicken breast, rice" # raw user text, kept for audit
items:
- { name: eggs, qty: 3, calories: 210, protein_g: 18 }
# ...
---
Dinner — 3 eggs, oatmeal, a chicken breast, rice
Logical-date rule
Timezone America/Toronto; a "day" runs 6:00am → 6:00am, so a meal logged at 12:30am belongs to
the previous calendar date. Conversion goes through the IANA timezone (DST-aware), never a fixed
offset. An explicit date passed by the user is used verbatim.
Tools
| Tool | Kind | Purpose |
|---|---|---|
log_meal |
write | Log a meal as one nutrition entry; totals derived from per-unit macros × qty. |
log_weight |
write | Log a bodyweight measurement (weight_lbs, optional waist/body-fat/notes). |
edit_meal |
write | Correct a nutrition entry by its key; recomputes totals. date moves the entry. |
edit_weight |
write | Correct a bodyweight entry by its key. date moves the entry. |
get_entries |
read | Fetch entries in a date range as compact JSON. Omit type for all types; pass it to narrow. |
delete_entry |
read | Delete an entry by its key + type (delete carries no fields, so it stays generic). |
get_daily_totals |
read | Per-day nutrition macro rollups for trend questions. |
Writes are typed (one log_X/edit_X per entry type, so each tool's schema shows exactly that
type's fields); reads are generic over the shared envelope. New entry types are added as an
adapter in src/registry.ts — see the Adapter interface (type, folder,
tags, zod input shapes, toBody/applyEdit/bodyLine).
Stack
- Cloudflare Workers + TypeScript
- Cloudflare R2 (
env.BUCKET) for storage - MCP over streamable HTTP, built on the agents SDK
- GitHub OAuth via @cloudflare/workers-oauth-provider, gated to an allowlist of GitHub logins
- Obsidian + Dataview for the dashboard (frontmatter properties, not inline fields)
Development
npm install # installs deps and the lefthook pre-commit hook
npm run dev # wrangler dev (local R2 via Miniflare)
npm run check # lint + format check + type-check
npm run lint:fix # apply lint fixes
npm run format # apply formatting
Tools can be exercised locally against the /mcp endpoint with the
MCP Inspector or any MCP client.
Deployment
Bindings (R2 bucket life-hub, KV namespace OAUTH_KV) are declared in wrangler.jsonc.
Secrets (set via npx wrangler secret put <NAME>):
GITHUB_CLIENT_ID,GITHUB_CLIENT_SECRET— from a GitHub OAuth App whose callback ishttps://<worker-url>/callbackCOOKIE_ENCRYPTION_KEY— e.g.openssl rand -hex 32
npx wrangler deploy
Then add https://<worker-url>/mcp as a custom connector in Claude and authenticate via GitHub.
Only GitHub logins in ALLOWED_USERNAMES (in src/index.ts) can use the tools.
Установка Life Hub
У этого сервера нет опубликованного пакета — он собирается из исходников. Открой репозиторий и следуй инструкции в README.
▸ github.com/mkadiya20/life-hub-mcpFAQ
Life Hub MCP бесплатный?
Да, Life Hub MCP бесплатный — установка в пару кликов через Unyly без оплаты.
Нужен ли API-ключ для Life Hub?
Нет, Life Hub работает без API-ключей и переменных окружения.
Life Hub — hosted или self-hosted?
Доступен hosted-вариант: Unyly запускает сервер в облаке, локальная установка не обязательна.
Как установить Life Hub в Claude Desktop, Claude Code или Cursor?
Открой Life Hub на unyly.org, выбери вкладку своего клиента (Claude Desktop, Claude Code, Cursor) и нажми Install — конфиг сгенерируется автоматически, без правки JSON.
Похожие MCP
Fetch
Web content fetching and conversion for efficient LLM usage.
AWS KB Retrieval
Retrieval from AWS Knowledge Base using Bedrock Agent Runtime.
автор: modelcontextprotocolSpring AI MCP Server
Provides auto-configuration for setting up an MCP server in Spring Boot applications.
llm-analysis-assistant
A very streamlined mcp client that supports calling and monitoring stdio/sse/streamableHttp, and can also view request responses through the /logs page. It also
автор: xuzexin-hzCompare Life Hub with
Не уверен что выбрать?
Найди свой стек за 60 секунд
Автор?
Embed-бейдж для README
Похожее
Все в категории ai
