Skip to main content
Glama
README.md
# Life Compass

An evidence-based reflection, goal and habit coach that runs inside **your own AI**.

Life Compass guides you from reflection to 1–3 goals to tiny daily habits, then keeps you going with short reviews. It's grounded in behavioural science: values work (ACT), goal setting, WOOP, implementation intentions, tiny habits, COM-B, habit-strength tracking and kind lapse handling.

It's an [MCP](https://modelcontextprotocol.io) server: you add it to an AI app you already use, and your AI becomes the coach. **Your plan is a file on your computer. Nothing is sent anywhere else.**

## Install in one click (Claude Desktop)

1. Download **[life-compass.mcpb](https://github.com/naya85/life-compass/releases/latest/download/life-compass.mcpb)**
2. Double-click it, or drag it into Claude Desktop (Settings → Extensions), and click **Install**
3. Start a chat and say **"Start my Life Compass"**

No Node.js or settings needed: Claude Desktop runs it for you.

## Other AI apps

### What you need

- [Node.js](https://nodejs.org) 18 or newer
- An AI app that supports MCP, for example Claude Desktop, Claude Code, Cursor, or LM Studio (for local models)

### Install

Add it to your AI app. It downloads and runs with `npx`, so there's nothing to clone.

**Claude Desktop**: Settings → Developer → Edit Config, then add:

```json
{
  "mcpServers": {
    "life-compass": { "command": "npx", "args": ["-y", "life-compass"] }
  }
}
```

**Claude Code**:

```bash
claude mcp add life-compass -- npx -y life-compass
```

**Cursor, LM Studio and other MCP apps**: add a server with command `npx` and arguments `-y life-compass`.

Restart the app.

## Use it

Start a chat and say **"Start my Life Compass"** (or pick the `life_compass` prompt if your app shows prompts). Your AI will:

1. **Reflect** with you, one question at a time (about 15 minutes)
2. Help you **choose** 1 to 3 goals tied to your values
3. **Plan** milestones and tiny habits with cues and if-then plans
4. Help you **act**: say "I did my walk" and it logs it
5. Run short **reviews**: "weekly review", "monthly check", "quarterly review"

Want calendar reminders? Say "export my reminders" and import the `.ics` file it creates.

## The dashboard

A visual Compass for your plan: rings for each goal, a mountain that grows with your progress, one-tap habit tiles, what's coming up, goals with milestones, and wins.

**No AI needed.** The dashboard can take you through the whole method on its own: reflect one question at a time, choose goals with WOOP, break them into milestones, build tiny habits with a cue, and do the weekly, monthly and quarterly reviews. Use the **+** button, or **Start reflecting** on a new plan.

```bash
npx -y -p life-compass life-compass-ui
```

Or ask your AI to "open my dashboard". It runs on your computer only, at a private link with a random key; other websites can't read or change your plan. Changes you make there and through your AI show up in both.

![Dashboard](docs/dashboard.jpg)

## Where your data lives

`~/.life-compass/plan.json` (change it with the `LIFE_COMPASS_DIR` environment variable). It's plain JSON you can read, back up or delete. A one-step backup is kept as `plan.json.bak`.

## What's inside

| File | What it is |
| --- | --- |
| `src/method.js` | The method: rules, questions and steps, each with the reason it's there |
| `src/coach.js` | Phase, what's due, capacity limits, habit strength (SRBAI), lapses |
| `src/store.js` | Local plan file, written safely |
| `src/index.js` | The MCP server: 21 tools, the `life_compass` prompt, and the method and plan as resources |
| `src/web.js`, `src/web/` | The local dashboard and its small, key-protected API |
| `test/` | End-to-end and dashboard security tests (`npm test`) |

## Not therapy

Life Compass is a self-coaching tool. If you're in crisis or struggling with your mental health, please reach out to a professional or your local emergency number.

## License

MIT

TDQS

B3.3/5.0

Scored across 21 tools

Disambiguation4/5

Most tools are clearly distinct: get_status/get_method/get_plan differ by what they return, and log_habit (record one occurrence) vs save_habit (configure/shrink/pause) is well separated by their descriptions. Minor overlap remains between save_baseline and quarterly_review (both record life-area ratings) and between save_milestones and tick_milestone, but descriptions give clear guidance on which to use when.

Naming Consistency4/5

The dominant pattern is verb_noun (get_status, save_goal, log_habit, tick_milestone, add_win, open_dashboard, export_reminders, record_decision), used consistently across the set. The review tools break the pattern as bare noun names (weekly_review, monthly_check, quarterly_review), and values_card_sort is noun-only, which are minor deviations but still readable.

Tool Count4/5

At 21 tools the surface is on the heavier side, but each maps to a concrete step of a multi-phase method (values, baseline, capacity, goals, milestones, habits, reviews, wins, dashboard, export). Nothing looks redundant, though the eight save_* tools reflect notable granularity that a heavier server might consolidate.

Completeness4/5

Coverage is broad: reading plan/status, all four planning phases, goal and habit lifecycle (including pause/graduate/drop via save_habit status), reviews, and export. The main gap is explicit goal deletion/completion and win removal, forcing agents to improvise rather than a dedicated operation.

Maintenance

ActivityMaintained
ResponsivenessNo recent community issues