LifeOS

A plain-text vault for your life plus a dashboard that renders it — driven by whatever AI assistant you already use.

Live demo — the dashboard running the demo persona, exactly what a fresh clone renders.

 · 

The deploy button points at this repo. If you fork LifeOS, change siddath/lifeOS to your own handle so it deploys your copy.

The dashboard running the demo persona — swap one config file and it renders your life instead.

---

What this is

Most "get organized" systems fail the same way: you spend an afternoon setting them up and never look at them again. LifeOS is built the other way around. Your life lives as plain Markdown and JSON in a Git repo you own; a static dashboard renders it; and an AI assistant does the tedious part — reading your messy notes and turning them into the structured files the dashboard reads.

There's no LifeOS server and no account. Your data sits in your repo, your browser's local storage, and — if you opt in — your own Notion and Vercel. Nothing phones home.

The one design decision everything hangs off: the framework ships no content, only paths and a contract. A single lifeos.config.json holds who you are; JSON Schemas in schemas/ describe every data file. Swap one mission.json and the whole dashboard re-renders around your current focus. Because the shapes are written down, any AI can read your notes and fill the system in — you don't need this specific tool or that specific model.

A fresh clone runs immediately as a demo persona ("Alex Rivera," a product engineer in Portland) so you can see the whole thing working before you touch a file.

Works with any AI

This isn't tied to one assistant. The instructions an AI needs live in AGENTS.md — a plain file that Claude, ChatGPT, Cursor, Copilot, Gemini, or anything else can read. Point your assistant at it and say "help me set this up." It recognizes the demo data as placeholder, asks for yours (or reads what you hand it — a paragraph, an Obsidian vault, a Notion export, a folder of notes), and maps it onto the schemas.

There's also a set of Claude Code skills (/setup, /daily-brief, /weekly-review, and friends) bundled under .claude/ for people who use Claude Code — a convenience layer, not a requirement. The schemas are what make it portable; the skills are one nice way to drive them.

What's in it

• One hero mission at a time. The dashboard centers a single focus with a live countdown, "the one thing right now," a week-by-week arc, and an evidence checklist. All of it renders from dashboard/mission.json; swap the file to swap the mission (about a ten-minute edit).
• Tasks and habits, live. A natural-language task box (#area P1 due:tomorrow), a weekly habit grid with streak history, saved in your browser — with optional best-effort two-way Notion sync so your phone and desktop stay close.
• A knowledge base about you. Searchable notes — strengths, watch-outs, preferences, one entry per life area. This is the raw material an AI reads to write your daily brief.
• The Anchor. A calm break-glass page: grounding steps, a few reframes, a seven-day reset. It's the part that assumes you're a person who spirals under pressure, not a productivity robot.
• Finance. Net worth, a monthly budget, and an optional broker view. Currency comes from your config.
• A four-step evening review. Two minutes: how the day felt, presence with the people you love, three small thanks, tomorrow's one thing.
• A warm, theme-aware UI. Light and dark, a mobile dock, reduced-motion support. Opinionated about being calm rather than loud.

Quickstart

git clone https://github.com/siddath/lifeOS.git
cd lifeOS

# Optional — make it yours. Skip it and you'll see the demo persona.
# Personalize in a PRIVATE copy (private fork/clone, or a new private repo
# from this template). The config holds identity + toggles, never secrets —
# but your name, areas, and life data still belong in a private repo, and
# committing the config there is what makes a deploy render as you.
cp dashboard/lifeos.config.example.json dashboard/lifeos.config.json

# Serve the dashboard — recommended, because file:// blocks the JSON fetches it needs
# (config, mission, tasks/habits seeds). No build step.
python3 -m http.server 8000      # then open http://localhost:8000/dashboard/

# Zero-server fallback (works, but some data may not load under file://):
open dashboard/index.html

Then open your AI assistant in this folder and tell it to read AGENTS.md and help you get started. It'll ask about you (or take whatever notes you already have), write your lifeos.config.json, and fill in your first mission, tasks, and knowledge base. If you use Claude Code, /setup does the same thing as a guided interview.

When you want it on your phone, deploy the dashboard/ folder with the button up top. It runs fine with zero environment variables — Notion sync just stays dormant until you add keys. Full deploy notes: docs/deploy.md.

Architecture

flowchart TD
    U[You] -->|any AI assistant| C
    U -->|edit / read| V[Markdown vault<br/>GitHub · Obsidian]
    U -->|browse| D[Dashboard<br/>static · dashboard/]

    subgraph Contract["The contract"]
      CFG[lifeos.config.json]
      SCH[schemas/*.schema.json]
    end

    C[AI assistant<br/>+ optional Claude skills] --> CFG
    D --> CFG
    V --> SCH
    C --> SCH

    D -->|edits queue| API[api/ serverless<br/>notion-sync · notion-pull]
    API -->|2-way sync| N[(Notion)]

    C -.MCP.-> G[Google Cal / Gmail / Drive]
    C -.MCP.-> B[Broker]
    V -.-> GH[(GitHub repo)]

    classDef dashed stroke-dasharray:4 4;
    class G,B dashed;

Loading

lifeos.config.json and schemas/ are the seam everything reads. The dashboard is static. The only backend is a pair of serverless functions you deploy for Notion sync — everything else is optional and connects over MCP. Longer write-up: docs/ARCHITECTURE.md.

Bring your own data

Import is prompt-driven against the schemas, not a rigid importer — so it works with whatever AI you like and whatever shape your notes are in. Four ways in:

Path	What happens	Effort
Just talk to your AI	Point it at AGENTS.md, answer a few questions, and it writes your config, mission, tasks, and knowledge base against the schemas.	~10 min
Paste your notes	Hand it a brain-dump, an old journal, a task list — messy is fine. It maps them onto the schemas and hands back valid data files.	~10 min
Notion	Build two small Notion databases from the property tables in the guide, set four env vars, pull — the dashboard hydrates from your Notion databases and syncs both ways.	~10 min
Existing tools over MCP	Let the AI read your calendar and notes over MCP and bootstrap the vault from what's already there.	~20 min

Details, including the demo persona and how the AI recognizes it as placeholder: docs/onboarding.md. The placeholder convention itself is in DEMO_DATA.md.

Make it yours

LifeOS ships a life, not a template of one — so the point isn't to fit yourself into the demo's shape, it's to rewrite the shape around you. Nothing here is fixed:

• Areas are whatever your life is actually made of. The demo uses career / health / relationship / fieldnotes / pottery / finance…. Yours might be phd / climbing / caregiving / freelance / faith. Edit the areas list in dashboard/lifeos.config.json — codes, labels, emoji — and the task composer, the knowledge base, and Notion sync all follow.
• Turn off what you don't use. The modules toggles (finance, anchor, knowledge_base…) hide whole surfaces. No money to track? Set finance: false and it's gone — section, nav, and all.
• The mission is the one thing you're driving toward right now. Swap dashboard/mission.json — a job hunt, a thesis, a move, a first 10k — and the entire hero re-renders around it. When it's done, drop in the next one with a fresh slug and the progress resets clean.
• Currency, locale, greeting, theme all live in the config, so it speaks your money and your language.

You don't have to touch code for any of this. Point your AI at AGENTS.md, tell it what your life looks like, and it edits the config and data files for you.

One privacy rule to keep straight: this public repo stays the clean template. Your personal instance — config, mission, tasks, knowledge base — lives in a private repo (a private clone, or a new private repo created from this template). The config holds identity and toggles, never secrets, and committing it in your private instance is what makes a deployed dashboard render as you instead of Alex. Treat any hosted personal dashboard as public unless you've put access protection in front of it.

Connectors

Connector	What v1 gives you	Setup
Vercel	Deploy button + vercel.json	One click
Notion	Optional two-way Tasks/Habits sync (best-effort); build two DBs by hand from the exact property tables in the guide	~10 min, guide
Google (Cal/Gmail/Drive)	Read access over MCP today; serverless sync is on the roadmap	Via MCP
GitHub	repo.url in config wires the vault-is-a-repo links	One line
Broker	Any broker's public MCP endpoint; the finance card also takes manual JSON	Via MCP
Something else	The connector contract: an env var, a data JSON, and an optional api/ function	—

Why it's shaped this way

LifeOS is opinionated in exactly one direction: it wants to move you from think → plan → improve the plan → think again to think once, then build, measure, improve. Two ideas carry most of the weight.

One hero mission at a time. Everything else is parked, not abandoned. Running every front at once is how you end up frozen; sequencing is the way out.

Evidence, not ideas. A week should end with something you can point at. The dashboard is built to make the artifact the win, not the to-do list.

It's meant for people who over-plan under pressure — who generate more meaning than they have containers for. The Anchor is where that has a calm home instead of a 1 a.m. spiral.

Docs

• docs/onboarding.md — the four ways to make it yours
• docs/ARCHITECTURE.md — how the pieces fit
• docs/deploy.md — put the dashboard on the web
• docs/connectors/README.md — the connector contract
• docs/connectors/notion.md — optional Notion two-way sync, step by step
• schemas/README.md — the data contract, file by file
• DEMO_DATA.md — the placeholder convention
• docs/INHERITANCE.md — how this template inherits functionality (never content) from its private upstream
• docs/ROADMAP.md — built now vs. planned

Contributing

Small, focused, private-by-default. See CONTRIBUTING.md and CODE_OF_CONDUCT.md. Don't commit personal data — CI runs a generic secret/PII scan on every PR.

License

MIT. Copying is the point, not the threat — build your own life on it.