Your AI agent plans the week. Persistent reminders keep you on it.
A local-first weekly routine, task list, and focus-block reminder app. One rmd binary gives you the CLI, the background reminder service, and a browser workspace.
Quick start · Workspace · Commands · Documentation · 中文
Tell Claude Code, Codex, or Antigravity about your life. The build-schedule skill interviews you and writes your weekly routine; plan-day fills today's focus blocks with your real tasks; suggest-tasks finds work in your email and calendar and leaves it on your board for you to accept or reject.
|
A background service sounds an alarm and shows a dialog at the start and end of every block until you respond. Deadlines, habits, and urgent tasks get their own nudges. | Your routine is TOML and your tasks and history are JSON, all on your computer. There is no account and no cloud service. The app makes no AI calls unless you connect a model to its optional Suggest button; agents talk to their own model provider. |
Healthy defaults, not just a full calendar
When your preferences leave gaps, build-schedule protects a few
evidence-informed defaults: at least 7 hours of sleep opportunity with regular
sleep and wake times, 150–300 minutes of moderate activity a week with
strength work on two or more days, short movement breaks in long sedentary
stretches, demanding work in your best-focus window, and daylight early in the
day. These are scheduling heuristics, not medical advice; your real
constraints always win.
- Watson NF et al. Recommended Amount of Sleep for a Healthy Adult. AASM consensus
- Sletten TL et al. The importance of sleep regularity. Sleep Health, 2023
- World Health Organization. Physical activity recommendations for adults. WHO
- Albulescu P et al. "Give me a break!" PLOS ONE, 2022
- Figueiro MG et al. The impact of daytime light exposures on sleep and mood in office workers. Sleep Health, 2017
The plugin's skills do the thinking, and rmd does the deterministic work
they need: rmd check validates your configuration and reports its paths,
rmd sync --json describes today's focus blocks and open tasks, and
rmd sync --apply saves today's plan. The reminder service (rmd daemon,
started at login) turns your files into alarms, and rmd web shows
everything in the browser.
The installer builds rmd from source, so you need
Rust first (cargo on your PATH). Then:
curl -fsSL https://raw.githubusercontent.com/AutoPhd-org/schedule-everything/main/install.sh | bashOr from a clone:
git clone https://github.com/AutoPhd-org/schedule-everything.git
cd schedule-everything
./install.shThe installer:
- builds
rmdwithcargo(from your clone, or from a copy of the repository it downloads) and puts it in~/SCHEDULE_MANAGEMENT. The first build takes a few minutes; - creates your config root in
~/.schedule_management/configwith a firstuser_config_0, and asks for any required settings that are missing; - registers the reminder service (a LaunchAgent on macOS, a systemd user service on Linux) and restarts it on the new binary when you upgrade;
- when LaunchX is installed,
registers the browser workspace with it from
~/SCHEDULE_MANAGEMENT, so LaunchX runs the installed binary rather than a source checkout.
Existing configuration is never overwritten. To update, run the installer again.
| Claude Code | OpenAI Codex | Google Antigravity |
|---|---|---|
claude plugin marketplace add AutoPhd-org/schedule-everything
claude plugin install schedule-everything@schedule-everything |
codex plugin marketplace add AutoPhd-org/schedule-everything
codex plugin add schedule-everything@schedule-everything |
git clone https://github.com/AutoPhd-org/schedule-everything.git
agy plugin install ./schedule-everything/plugins/schedule-everything |
Inside a Claude Code session you can run the same steps as
/plugin marketplace add and /plugin install. The installer prints these
commands too, and the workspace's Schedule page has copy buttons for them.
| Say | What happens |
|---|---|
| "Build my weekly schedule" | build-schedule asks about your work, constraints, energy, and habits, shows a plain-language summary, and only after you confirm writes settings.toml, odd_weeks.toml, even_weeks.toml, habits.toml, and profile.md. It validates them with rmd check and reloads reminders. |
| "Plan my focus blocks for today" | plan-day reads rmd sync --json, matches your most important tasks and nearest deadlines to today's pomodoro and potato blocks, confirms with you, and saves the plan with rmd sync --apply. |
| "Move my workouts to the evenings" | Changes become a new user_config_<n> version, validated with rmd check --dir and activated with rmd switch <n>. Earlier versions stay for rollback. |
| "Suggest tasks from my email and calendar" | suggest-tasks reads rmd suggest ls --json, looks through your connected apps for work you have not captured, and saves a few suggestions with rmd suggest add. They wait on the task board, marked as suggestions, until you add or reject each one. Run it on a schedule and new suggestions keep arriving. |
rmd webThe workspace listens on 127.0.0.1 only and accepts API calls only with the
session token of the tab it opens. It reads and writes the same files as the
CLI, and running rmd web again reuses the open tab. To use it from your phone
as well, see On your phone, too.
If you run your local services with LaunchX, register this repository with
lx add and the workspace becomes one of them: rmd web then hands over to
lx open rmd, and scripts and agents can read and change it over HTTP. Allow
it to use LaunchX's shared memory (lx memory access rmd write) and it keeps
your routine and upcoming deadlines on the memory's schedule page for your
other apps and agents, and Suggest reads your profile and preferences from it.
See LaunchX.
Five pages: Overview, Schedule, Routines & deadlines, Activity, and Settings.
J mode is the structured day: overdue, due-today, high-importance, and
habit counts, the current and next block, upcoming deadlines, and today's
timeline. P mode is for flexible days: alarms for selected events pause
and the page keeps only your tasks. Flip the J / P switch in the header or
run rmd mode j|p.
- Task board. One group per task type. The compact layout puts groups with more important work first; spread places them by workload, the heaviest in the middle. Groups keep their places while you work. Add a task inside its group, give it a finish-by time, and track it with a progress slider or a checklist of subtasks. Fold turns each group into a summary of its importance, progress, and size. Task types are added, renamed, and deleted right on the board.
- Suggestions from your agent. Tasks an agent suggests with
rmd suggest addwait in their type's group, dashed and labeled with who suggested them and why. They are not tasks yet: no reminder, count, or statistic includes them. Add to my tasks makes one a normal task, Adjust lets you change it first, and Reject turns it down so the agent does not suggest it again. Both can be undone. - Suggest. Paste an API key for Claude, or point it at any OpenAI-compatible service (OpenAI, OpenRouter, a local Ollama), in Settings → AI suggestions. Suggest on the board then asks that model for a few tasks you are likely to need, from your open tasks, deadlines, today's schedule, habits, recent work, and profile, and, under LaunchX with your permission, the profile and preferences in its shared memory. Its ideas wait on the board as suggestions, like an agent's.
- Timers. Dials above the board count down to timed tasks due by tonight (overdue ones count up) and to deadlines in their last 24 hours. One bar shows the progress of today's timed work, optionally weighted by importance.
- Daily overview. What you finished today and when, compared with yesterday and your usual day, plus your streak, habits, and the open work left.
- Workload. Every open task counts its importance out of 10 times the share still left (an importance-8 task counts 0.8 untouched and 0.4 half done). The theme's colors follow the total, from All clear through Light, Steady, and Busy to Heavy. Settings sets where the levels start, or turns the workload theme off.
Today's timeline shows each block with its planned task. The page also has the plugin install commands with copy buttons, example prompts, and a button that clears today's plan.
Every deadline sits on a year calendar. Click a day to add one, make it repeat weekly, every two weeks, or monthly, and pick one of eight marker colors. Recolor or edit it from the day card, or drag its chip onto another day to reschedule. Repeating deadlines show every occurrence and roll forward on their own. Turn on AI conferences to add submission deadlines, rebuttal periods, and conference dates for NeurIPS, ICLR, ICML, CVPR, ACL, ICRA, and other major venues, in your time zone. Today's habits live on the same page.
Everything you finished, cancelled, or let go grows along a scrollable timeline tree, one lane and color per task type. Prefer cards? The gallery gives each task artwork in its type's color. Filter by outcome or type, search, or click a day in the rhythm grid. Hover over any task to see when it started, how long it took, and how that compares with your other tasks.
![]() |
![]() |
| Every task has a story. Click a task for its timeline and pace. | Your week in color. The weekly overview compares this week with last week and your recent average. |
A plugin is a task category with its own page and tracking. Its tasks are ordinary tasks of its type, so they show on the board and in Activity, and you choose in Settings whether it gets its own page, its own metrics in Activity, a card in the timer row, and whether your AI agent may read it to suggest tasks. The built-in Fitness plugin has a body figure that lights up the muscles you trained this week, a workout builder with adjustable units (sets, reps, weight, rest), a live session with a rest timer, personal records, and body weight. Build your own trackers for practice time, new words, pages read, or anything else, each with weekly goals. Activity can be viewed by category, so your fitness history and metrics stand on their own.
Plugins; board, workload, and progress preferences; reminder preferences; task and
deadline types; the model Suggest asks; under LaunchX, what the workspace may do
with your shared memory; and an editor for your configuration files that checks every
save with the same rules as rmd check and refuses to overwrite newer edits.
![]() |
![]() |
![]() |
Run rmd web --phone and scan the code it prints with your phone's camera. The
phone then uses the workspace on your computer, so it works on the same tasks,
plan, and history while you keep working at the computer. Nothing is copied or
synced, and the phone stays signed in.
rmd web --phone # turn phone access on and show the pairing code
rmd web --no-phone # turn it off and sign every phone outThe phone needs to be on the same Wi-Fi as the computer, or on Tailscale from anywhere, and the computer needs to be awake. With phone access on, the server also listens on your network and accepts requests addressed to the computer's IP address that carry the phone's token. The link travels unencrypted over your local network, so use phone access on networks you trust or over Tailscale. The reminder service starts the workspace at login while phone access is on. Alarms still sound on the computer.
Light, dark, or system theme, remembered per browser. Motion follows your system's reduced-motion setting.
# odd_weeks.toml (even_weeks.toml has the same shape)
[wednesday]
"09:00" = "pomodoro" # a [time_blocks] name: alarm at start and end
"14:00" = { block = "meeting", title = "Advisor 1:1" }
[common] # every day; a weekday entry at the same time wins
"12:30" = "lunch_time" # a [time_points] name: one reminder
"20:00" = "potato"Untitled pomodoro and potato blocks are the focus slots a daily plan can
fill. A plan changes only today; your weekly files never change.
| Command | What it does |
|---|---|
rmd web [--port N] [--no-open] [--phone | --no-phone] |
Open the browser workspace, reusing a running server and its tab; --phone lets your phone use it too |
rmd serve [--port N] · rmd state |
Run the workspace server in the foreground (LaunchX runs this), or print everything the workspace shows as JSON |
rmd status [-v] |
What is happening now and next, with today's planned task titles |
rmd add "task" 8 [type] [days] |
Add a task with importance 1–10, an optional type ID, and optionally postpone its urgent alarms by some days |
rmd ls [-t TYPE] |
List tasks by importance, optionally of one type |
rmd rm · rmd cancel · rmd drop |
Remove tasks by description or ID: finished (counts as done), added by mistake, or given up |
rmd history [n] |
Recent finished, cancelled, and dropped tasks |
rmd ddl · rmd ddl add <event> <M.D> [--repeat weekly|biweekly|monthly] · rmd ddl rm <event>… |
List, add, and remove deadlines, including repeating ones |
rmd track [id…] |
Record today's habits (asks about each one when no IDs are given) |
rmd sync [--json | --apply FILE | --clear] |
Show, save, or clear today's plan; --json and --apply are what plan-day uses |
rmd suggest [ls [--json]] · rmd suggest add "task" 8 [--type T] [--reason …] · rmd suggest rm <id>… |
List, add, or withdraw tasks suggested by an AI agent; you accept or reject them on the task board |
rmd check [--dir DIR] [--json] |
Validate a config directory and report its paths |
rmd mode [j|p] |
Show or switch between J and P mode |
rmd switch <n> |
Activate user_config_<n> and reload reminders |
rmd update · rmd stop |
Reload reminders (after a git pull when the config root is a git checkout), or stop them |
rmd settings · rmd edit [file] |
Edit settings.toml in a terminal UI, or open a config file in your editor |
rmd view · rmd report weekly|monthly [-d DATE] |
Export a schedule PDF or a productivity report |
rmd completion [bash|zsh|fish|tcsh] |
Print a shell completion script |
~/.schedule_management/config/
├── user_config_0/ # one complete version of your routine
│ ├── settings.toml # time blocks, time points, reminder times, task types
│ ├── odd_weeks.toml # schedule for odd ISO weeks
│ ├── even_weeks.toml # schedule for even ISO weeks
│ ├── habits.toml # daily habits
│ ├── ddl.json # deadlines (with optional repeats and marker colors)
│ └── profile.md # what your agent knows about you
├── user_config_1/ # a later version made by build-schedule
├── synced_schedule.toml # today's plan
└── tasks/ # tasks, history, habit records, mode
Edit any file by hand, in the workspace's configuration editor, or with
rmd settings, and run rmd check after manual edits. REMINDER_CONFIG_DIR
points rmd at a different config root.
- Introduction
- AI agent plugins
- Quick start
- Installation
- Browser workspace
- Configuration
- CLI reference
| Path | What it holds |
|---|---|
crates/rmd |
The rmd binary in Rust: CLI, reminder service, local web server, PDF reports |
web |
The browser workspace in TypeScript, bundled by Vite into crates/rmd/web-static |
plugins/schedule-everything |
The agent plugin and its build-schedule, plan-day, and suggest-tasks skills |
config |
Templates the installer copies into a new config root |
documentation |
The Docusaurus documentation site |
tests/golden |
Recorded CLI and browser API cases that cargo test replays |
cargo build # target/debug/rmd
cargo test # unit tests and the golden cases
cargo fmt --all --check
cargo clippy --all-targets -- -D warnings
bun install
bun run test # browser tests (Vitest; plain `bun test` is Bun's own runner)
bun run build # rebuild crates/rmd/web-staticJavaScript dependencies are managed with Bun (bun.lock);
the tools themselves run on Node.js 20 or newer. The built frontend is
compiled into the rmd binary, so commit crates/rmd/web-static with
frontend changes and rebuild rmd to serve them. To work on the frontend
with live reload, run rmd web --port 8765 --no-open and bun run dev.
The golden cases were recorded from an earlier Python implementation that
rmd replaced, and they pin its CLI output, browser API responses, and data
files. See tests/golden/README.md and
crates/rmd/PORTING.md.
Distributed under the MIT License. See LICENSE.









