pixari / OpenTodo

OpenTodo 
 A to-do app that runs on your own server. 
Plan your day, keep projects in order, set reminders and share lists with your family or your team. Works on your phone and your computer, even offline. No monthly plan per person, and your tasks stay with you. 
 Try the demo (no sign-up) ·
 Install in 2 minutes ·
 Read the docs ·
 Website 
 If OpenTodo looks useful, a star on GitHub helps other people find it, and a coffee helps me keep building it. 
---
## Try it in 30 seconds
- **No install:** open the demo and press **Try the demo**. You get a private copy with sample data, deleted after a few hours.
- **On your machine or server:**
 [code] 
 Open ; the first visit creates your account. No configuration needed. Coolify, Unraid, Compose and plain binaries are in Quick start.
Questions, ideas and bug reports are welcome in issues. To talk to other people who use OpenTodo, join the Discord server.
## Why OpenTodo
- **No subscription, no lock-in.** Free software under the AGPL. Nobody can paywall your task list or shut it down with the company.
- **Works offline, on every device.** Each phone and computer keeps a full copy of your data (local-first), so the app works with no connection at all. The server syncs, backs up and shares; it is never in the way of an edit.
- **Your data, one folder.** Everything lives in one `/data` volume around a single SQLite file. Export to JSON at any time, import it (or a CSV, or another app's export) on any instance, and schedule encrypted backups to disk, S3 or WebDAV.
- **Tiny footprint.** One container, one static binary of about 23 MB with the web app inside, around 20 MB of memory when idle. Runs on a Raspberry Pi or the cheapest VPS. No database server, no Redis, no mandatory configuration.
- **Private by design.** No telemetry, no analytics, no phone-home, no third-party scripts or fonts. The server only contacts what you configure (see Privacy).
- **Bring your own AI, if you want it.** Plan and change tasks by talking to the assistant in the app (Anthropic, OpenAI, Gemini or a model on your own network), connect the assistant you already use over MCP, or let a coding agent drive the `ot` CLI. Off by default, every change recorded and undoable, no AI vendor built in. See Use AI with OpenTodo.
- **Keyboard-first.** `q` to capture, `j`/`k` to move, `c` to complete, `1`–`4` for priority, `g` `t` for Today, `?` for the full list, and optional Vim keybindings.
- **Built in the open with specs.** Every feature starts as an OpenSpec change under `openspec/`, so behaviour is documented before it is built.
## Status
OpenTodo is at **0.4**: every feature on the roadmap is in, covered by automated tests, but some integrations still need real-world testing (CalDAV with specific apps, single sign-on providers, real phones). If you try it, tell us how it went in the call for testers.
**How it's built:** every feature starts as a written spec (OpenSpec), and much of the code was written with AI coding agents working f