Delivers outbound WhatsApp messages from the board: heads-up reminders before meetings and blocked work time (with the linked task's next step), scheduled digests, and device pairing codes. The agent receives the user's WhatsApp requests and files them into the board via MCP.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Workboard תוסיף לקניות חלב וביצים"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A self-hosted task board for one person or a household — that your AI agent runs for you.
Work, personal and home in one place · a schedule that taps you on WhatsApp before every meeting · shared shopping lists that know the supermarket aisles · Hebrew-first, RTL, built like an iPhone app.
Most task apps assume you'll sit down and type. Workboard assumes you won't. You tell your agent on WhatsApp "קבעתי עם הספק לשלישי ב־10" or "תוסיף לקניות חלב וביצים", and it files the task, books the slot, reminds you ten minutes before, and keeps the board honest. When you do open the app, it's a calm, fast PWA that feels at home on an iPhone and shows the one thing that matters next.
Related MCP server: SoloOS MCP Server
🗂 Contexts, not folders Work, personal and a shared home space. One tap switches the whole app — colour, tabs, what Today shows — or let it follow your work hours automatically.
👥 Built for two (or more) Each person has private spaces and their own agent; the household shares a home space. Shared tasks have an owner (עליי · עליה · שנינו), and nobody can see the other's private tasks — enforced on the server.
🗓 A schedule that reminds you Meetings and blocked work time, coloured by the part of life they belong to. A WhatsApp heads-up before each one (15 min before meetings, 10 before work blocks), with the linked task's next step.
🛒 Shopping that knows the store Supermarket, pharmacy, house, personal and wish lists. Items sort themselves into aisles (and learn your fixes), a big-target in the store mode keeps the screen on, and on a wish list you can quietly mark "I'll buy this" — the one who wished never sees it.
🤖 Agent-native (MCP) 13 typed tools over MCP. Hebrew dates are resolved on the server ("חמישי שאחרי", "מחר ב־7:30 בערב"), batch writes report per item, every change is undoable.
📱 Feels native Installable PWA, offline read-only snapshot, live updates, swipe actions with spring physics, light & dark, Dynamic Type, full RTL, and copy that addresses each person in their own grammatical gender.
Link each person's work space to a team board: tasks you own or review there appear here, edits go back, and a personal task can be put on the team board. docs/team-board.md
Action items from a Pocket AI recorder's meeting summaries arrive as proposals awaiting your approval, with the summary sent to WhatsApp. docs/pocket.md
git clone https://github.com/TalEps77/Personal-Workboard.git
cd Personal-Workboard
# Your instance lives outside the repository: config, secrets and data.
mkdir -p ~/workboard-instance/data
sudo chown -R 1000:1000 ~/workboard-instance/data # the container runs as uid 1000
cp examples/.env.example ~/workboard-instance/.env # edit: hosts, time zone, tokens, and set
# WORKBOARD_DATA_DIR / WORKBOARD_ENV_FILE to ABSOLUTE paths
cp examples/people.json ~/workboard-instance/data/ # optional — for more than one person
docker compose --env-file ~/workboard-instance/.env up -dOpen it through an HTTPS reverse proxy (Caddy, Tailscale serve, nginx — see deploy), pair your
phone with the code that arrives on WhatsApp (no agent connected yet? the code is in docker logs workboard), and
Add to Home Screen.
Just want to look around? npm run dev starts a local board with sign-in disabled at http://127.0.0.1:4173.
you ──WhatsApp──▶ your agent (e.g. Hermes) ──MCP tools──▶ Workboard ◀──PWA── your phone
▲ │
└── signed webhook: reminders, digest, pairing codes ──┘Workboard — one Node 22 process, no dependencies, files as the database (each task is a readable Markdown file), a single lock and atomic writes, per-person scoping of everything it serves.
Your agent — talks to you and calls the tools. Workboard does the date math, the reminders and the delivery.
Your phone — the dashboard: Today, the schedule, all tasks, projects, shopping.
MCP setup, delivery webhook, one agent per person, the agent skill | |
Every tool and parameter (generated from the code) | |
Environment variables, | |
People, spaces, visibility, shared tasks, reminder routing | |
Reverse proxy, backups, the safe auto-update pipeline | |
Modules, data model, live updates, concurrency | |
Threat model and controls · report a vulnerability | |
Tokens, components and rules — readable by people and by models |
Self-hosted instances can follow main without trusting every push: deploy/update.sh
deploys a commit only if it is a fast-forward, its GitHub checks are green, the unit tests pass again
locally in a network-less container, and the new container comes up healthy — otherwise it rolls back.
Open apps pick up the new build by themselves. Details in docs/deploy.md.
npm run dev # local board, sign-in disabled, loopback only
npm test # unit + API tests (node --test)
npm run test:browser # Playwright suites: single-user, two people, iOS zoom checks (PW_DIR=/path/to/playwright)
npm run docs:tools # regenerate docs/tools.md after changing an MCP toolNo npm install needed — the app has zero runtime dependencies. See CONTRIBUTING.md.
The interface, the agent replies and the date vocabulary are Hebrew (English date words work too), and the whole UI is RTL-first. The code, the docs and the API are in English. Translations are welcome.
This server cannot be deployed
Task & board management for AI agents + humans. Kanban, comments, digests via MCP.
Manage Superlist tasks and lists in plain language from any MCP-compatible AI agent.
Work management where AI agents are first-class members: tasks, projects, memory over hosted MCP
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
