WatchTower — Local-first activity tracking

Privacy-focused, local-only time tracking in the spirit of ActivityWatch. All data stays on your machine. No cloud, no telemetry.

What it is

WatchTower tracks what’s happening on your computer (active application, window title, optional browser tab, AFK, etc.) and stores everything in a local database. Data never leaves your device.

Primary goal, 03/05/26 (current)

A simple application with a lightweight interface that:

• Tracks activity on the user’s computer (active app, window title, optional browser tab, AFK).
• Shows a simple list of activity with basic filtering.
• Shows simple statistics only — e.g. “Cursor 3h today”, “Slack 45m”, “Chrome – GitHub 20m”, “Google Meet 1.5h”, “AFK 4h”. No charts or graphs in the first version.

Tech stack

• Interface / client: Next.js
• Storage: MongoDB (run locally)
• Environment: Docker (and Docker Compose) so the same setup runs on any platform

Local-first

All tracked data is stored only in a local MongoDB instance (e.g. in Docker with a named volume). No server outside the user’s machine. The design is “local DB”; the first implementation uses MongoDB.

Future: NPM package

This repository is intended to evolve into an NPM package so other parts of your software can depend on it (e.g. API client and TypeScript types). The Next.js app will be the reference implementation or a companion UI.

Design decisions

• Docker Compose: MongoDB (and optionally the full stack) runs via Compose. Data lives in a named volume so it persists and stays on the host. Environment-safe and local-only.
• Core as a local API: The core is a service that (1) accepts tracking events (app, window, duration, etc.) and (2) exposes read endpoints for “today’s list” and “simple stats” (aggregations by app/duration). The NPM package will expose a client SDK and types; other products can npm install and talk to the same local API.
• Tracking collectors: ActivityWatch-style “watchers” (desktop process for active window, browser extension for tabs) are on the roadmap. A small desktop agent or browser extension will send events to the local API (separate repo or sub-package later).
• Simple stats (v1): Store raw events (app name, window title, start/end timestamps, optional category). The API aggregates by day and by app/window and returns simple summaries. The UI only renders lists and these summaries — no charts in v1.
• Database: First implementation is MongoDB. The design could later support alternatives (e.g. SQLite for a zero-Docker, single-file option) while staying local-only.

Getting started

Prerequisites: Bun (optional) or Node.js LTS, Docker.

1. Clone the repository

   git clone https://github.com/SarkarShubhdeep/WatchTower.git && cd WatchTower

2. Run the app (choose one)

   Option A — Full stack in Docker, production build (no hot reload)

   docker compose up -d

   Open http://localhost:5800.

   Option B — Full stack in Docker with hot reload (good for development)

   docker compose -f docker-compose.yml -f docker-compose.dev.yml up

   The app runs in dev mode; code changes in app/ are picked up automatically. Open http://localhost:5800.

   Option C — MongoDB in Docker, app on your machine (also good for dev)

   docker compose up -d mongodb
   cp .env.example app/.env
   cd app && bun run dev

   Or from the repo root: bun run dev. If you use npm: cd app && npm run dev. Open http://localhost:5800.

3. Optional: Edit app/.env to override MONGODB_URI (only needed for Option C). The API route /api/health pings MongoDB when the app is running.

Stop: docker compose down (data persists in the volume). With Option A or B, both containers stop; with Option C, only MongoDB stops.

License

See LICENSE (MIT).