A planner for projects where you don't know what the next task is yet.
Built for research, design exploration, and other work whose durations are genuinely uncertain — where progress matters more than deadlines.
Why it exists · Distinctive features · Screenshots · Quick start · Status
Most project trackers — Jira, Linear, Asana, even Notion — were designed for scoped engineering work: you know what to build, the question is when. They treat every task as a known-duration block on a calendar.
That model breaks the moment your work is genuinely uncertain:
"I think this experiment takes a week. It might take three. Or I might realise on day two that the whole approach is wrong."
Researchers, PhD students, designers, independent makers, and anyone running multiple long-running exploratory threads will recognise the shape of this problem. The traditional response — fill in an estimated due date and pretend — generates anxiety without generating clarity.
Research Planner is built around the opposite premise: when a task's duration is unknown, progress is the signal, not a missed deadline. The product is shaped by three beliefs:
- Progress > deadlines. What you're currently doing and what's stuck is more useful than a Gantt-chart fiction.
- Capture > structure. A thought should reach the inbox in one second. Tags, estimates, and dependencies are optional.
- Context on re-entry. Opening a project after two weeks away should tell you where you left off — not show you an undifferentiated task list to re-read.
The first thing the app shows you — not a chronological backlog, but a reading of this moment: what you're doing, what's blocked, what you pinned for today. Includes:
- Daily check-in — a lightweight "since yesterday → pinned for today → blocked" glance, derived from your own task movement. A reflection, not a report card.
- Stuck-task detection — surfaces tasks that have been
doing>7 days orblockedat all. - Project re-entry briefings — when you open a project after >3 days, a card summarises what you left mid-flight (now matched on the project page itself).
- Capacity rail — live counter of your committed intensity vs. today's budget, plus a next-7-days load preview.
A note in the inbox in one keystroke — even mid-sentence in another field. No project, no fields, no estimate. Inline hashtags (#literature #q3) auto-extract on save. Everything lands in a dedicated Inbox page; triage later — or never. Optionally attach a capture to a specific task as context.
- Progress mode (default) — Flow board (kanban) + notes + artifacts. Hides Gantt, dependencies, and scheduling until they're actually useful. The view for "I'm exploring."
- Deadline mode — Uncertainty Lane, Gantt, scenarios, PERT-driven scheduling, critical-path, soft/hard due dates. The view for "this paper must be submitted by the 15th."
Switching is one click per project. Most projects live in progress mode forever.
Each project keeps a tab of its artifacts — the links, files, code, datasets, and notes it produced or referenced. The PDF, the DOI, the repo, the figure: attached to the project instead of lost in a folder.
Visualise tasks as PERT confidence cones (Optimistic / Most-likely / Pessimistic) instead of single-point bars. The lane shows the range you're operating in, and the critical path is highlighted across the cones — not pretending a 3‑week stretch is "exactly Tuesday." A classic single-line Gantt sits behind a second tab for when you want it.
Save the current schedule as a named scenario, then change something — defer a task, add capacity, drop a dependency — and overlay the new schedule against the saved one as a faint dashed reference. "What if I commit four hours a day instead of eight" becomes a one-click question.
A reflection page, not a report card. Asks "what surprised you this week" — not "did you hit your sprint goals." Surfaces dormant projects (no activity >14 days), capture counts, what's still blocked.
Built on a Fastify WebSocket layer with presence (who's looking at which project), event broadcasting (project/task/note changes propagate to other connected sessions), and proper per-workspace role-based access. The default deployment is single-user / loopback-only; multi-user mode is a one-env-var flip.
Bundled as a Tauri app with a Bun-compiled API sidecar — a single signed .app that runs offline, ships with a seeded SQLite database, and has no external service dependency. (Linux and Windows builds are mechanical to add.)
Screenshots live in
docs/screenshots/. See the capture brief for what each image is meant to convey.
| Top-of-Mind | Flow Board |
|---|---|
 |
 |
| Uncertainty Lane | Weekly Review |
|---|---|
 |
 |
Requirements: Node 20+, npm 10+. Optional: Bun (only if building the Tauri sidecar), Rust toolchain (only if building the native app).
# Clone + install
git clone https://github.com/xtao-sh/research-planner.git
cd research-planner
npm install
# Seed the local SQLite + start both web + server in dev mode
cd apps/server && npx prisma migrate deploy && cd ../..
npm run dev
# → web at http://localhost:5173
# → API at http://localhost:4000The dev script starts the Fastify API on :4000 and the Vite dev server on :5173. The default deployment is single-user, loopback-only (no auth required, no LAN exposure). See apps/server/.env.example to flip on multi-user / network mode.
cd apps/web
npm run tauri:build # ./src-tauri/target/release/bundle/macos/Research Planner.appThe script Bun-compiles the API into a single binary sidecar (~60 MB), seeds a SQLite database, and embeds both into the .app bundle. First launch copies the seed database to ~/Library/Application Support/com.researchplanner.desktop/.
| Layer | Choice |
|---|---|
| Frontend | React 18 · Vite · TypeScript · @dnd-kit · i18next (EN / 中文) |
| Backend | Fastify · Prisma · SQLite (default) / Postgres · WebSocket (in-house presence + broadcast) |
| Desktop | Tauri 2 · Bun-compiled sidecar |
| Testing | Vitest (web · server · scheduler suites) · Prisma test fixtures |
| Type system | Shared @rp/shared package for cross-stack types · zod where it earns its keep |
The project is alpha but daily-usable. The author uses it as the primary planner for real research work. It's not production-deployed for paying users; expect rough edges, some unimplemented affordances, and breaking-change migrations between minor versions until 1.0.
What works today:
- ✅ Full task + project + note CRUD with realtime sync
- ✅ Flow Board (progress mode) · Task List, Uncertainty Lane & Gantt (deadline mode) · Artifacts tab
- ✅ Quick Capture (⌘⇧N) · Inbox · Top-of-Mind · Search (server-side) · Command Palette (⌘K)
- ✅ Scenarios + PERT scheduling + critical-path
- ✅ Multi-workspace with role-based access (owner / admin / writer / viewer)
- ✅ Self-contained
.appbuild (macOS); structure ready for Linux/Windows - ✅ Dark mode · EN/中文 locale switch
- ✅ Accessible focus traps, per-route error boundaries, toast system
What's deliberately not built yet:
- ⏳ Mobile-first responsive layouts (the app works on tablet, but is desktop-first by design)
- ⏳ Email integration / reminder loop
- ⏳ Public sharing / read-only project links
- ⏳ Native Linux / Windows desktop builds (Tauri config is portable but untested)
PRs welcome — see CONTRIBUTING.md for setup, commit style, and the (deliberately small) scope guidelines.
Bug reports are most useful when they include: the workspace + project mode, the exact action sequence, what you expected, what happened. Screenshots help.
- Product narrative & PRD:
docs/PRD-Research-Planner.v2.zh-CN.md(Chinese; the source-of-truth for product direction) - Architecture:
docs/TECH-ARCH-Research-Planner.zh-CN.md - Screenshot capture brief:
docs/screenshots/README.md
The MIT license keeps copyright with the author while granting permissive use rights. You may use, fork, modify, and redistribute, including commercially — please retain the copyright notice.
If this tool fits how your brain works, I'd love to hear about it.