Move Along：AI 任务管理应用，把想法一键拆解为可执行步骤

Move Along

A task-management app that moves ideas into action and reduces decision fatigue. Add a task and AI immediately breaks it into specific, actionable steps — with real links, prices, and decisions surfaced — so you can act instead of plan. Postpone, complete, lock to a deadline, chain into series, prioritize, or hand a step to an AI agent, each with a single click or keypress.

Built deliberately simple so you can fork it and make it your own: a vanilla-JS single-file frontend, a small Express API, and SQLite with zero native dependencies. No build step, no framework, three npm packages.

Features

• 30-day calendar board with horizontal scroll; overdue tasks spill forward to today automatically
• AI-generated subtasks — adding a task opens a pane of up to 10 concrete steps (Anthropic API; falls back to mock steps without a key)
• Series (task chains) — drag a task onto another to sequence them; countdown circles, hover threads, cascading reschedules, ✂️ to unlink
• Deadlines — lock a task to a date; locked tasks never spill, and overdue locked work pulls the board back to it
• Priorities — hover a task, press 1/2/3 for red urgency marks; the list re-sorts
• Hyperlinked tasks — URLs in task text become clickable links
• Day counters — see how many days each task has been pushed forward
• Multi-user assignment — assign tasks or single steps to teammates or AI agents
• Projects — as many boards as you want per user, plus a "Main" page grouping everything by project

Quick Start

cd server
npm install
npm start          # or: npm run dev  (auto-restart on change)

Open http://localhost:3000. That's the whole setup — the server serves both the API and the frontend, and the database file creates itself on first run.

Environment Variables

All optional:

Requires Node.js ≥ 18 (native fetch).

Deploying

Any Node host works. On Railway/Render/Fly:

1. Set the start command to npm start in server/.
2. Attach a persistent volume and point DB_PATH at it (e.g. mount at /data, set DB_PATH=/data/movealong.db). Without a volume, every deploy wipes the database.
3. Optionally set ANTHROPIC_API_KEY for real AI subtasks.

Cost warning: there is no authentication. If you deploy publicly with an ANTHROPIC_API_KEY, set AI_ACCESS_KEY too — then only browsers where you've entered that key (Shift+Click the 🧠 toggle, once) can trigger billed AI calls; strangers get free mock steps. The per-IP and daily rate limits are a second layer capping the worst case even for key holders.

Project Structure

server/public/index.html   Frontend — single file, inline CSS + JS, no build step
server/src/server.js       Express API
server/src/db.js           SQLite layer (sql.js — pure JS, no native compilation)
server/src/ai.js           Anthropic API client for subtask generation
reference/                 API reference, data model, roadmap
CLAUDE.md                  Working notes for AI-assisted development on this repo

Design Choices & Known Limits

Simplicity is the point; these are deliberate trade-offs, fine for personal/small-team use:

• No authentication — boards are reachable by URL. Add auth before storing anything sensitive.
• sql.js keeps the whole database in memory and rewrites the file on change (atomically, once per request). Great up to tens of MB of tasks; if you outgrow it, swapping db.js to better-sqlite3 is the upgrade path.
• Single-file frontend — ~3,700 lines of vanilla JS with clearly sectioned CSS, state, rendering, and handlers. Easier to fork and read top-to-bottom than a framework build; not built for a large frontend team.
• No test suite — manual testing against http://localhost:3000.

License

MIT — fork it, strip it for parts, build your own. See LICENSE.