Can AI replace Skribbl?
The canvas is the easy half. React Flow plus xterm.js plus node-pty gets you panning, zooming and half a dozen live agent terminals in a weekend, and it is a genuinely nice way to watch a fleet work. The gaps are all in the boring layer underneath: knowing when an agent is actually blocked on you rather than just quiet, keeping sessions alive across a restart or a crash, and counting tokens accurately enough that the number can be trusted. Each of those is days of work on its own, and the weekend build reads as finished until the first agent silently dies behind a tab you were not looking at.
01What it costs
Checked Aug 14, 2026 · source: skribbl.dev.
| Plan | Monthly | Billed yearly | What you get |
|---|---|---|---|
| Founding Infinity | $29 | $8.25/mo | 3 concurrent agents; founding-plan access while licenses remain available. |
Hidden costs: The plan does not include Claude Code, model API or other agent-provider subscriptions; the trial requires a card from day 2 and charges on day 15 unless cancelled.
02Could AI build it for you?
The core job: Spawn real shells as draggable nodes on a pan/zoom canvas, run a coding agent in each one, and watch several of them work side by side.
What a working version needs:
- Node 20 and native build tools for node-pty
- A coding agent CLI already installed and authenticated
- tmux, if you want sessions to survive a restart
An honest kinda · the canvas is a weekend, the process supervision under it is not.
03What you'd give up
- hook-driven agent status, so a blocked agent looks the same as an idle one
- session survival across restarts and crashes
- an accurate token meter, cache reads price nothing like input tokens
- a signed and notarized Mac build, so Gatekeeper blocks what you ship to anyone else
Not for the canvas. For the layer that keeps twelve child processes honest: status that comes from the agent itself instead of scraped output, sessions that come back after a reboot, and a spend number accurate enough to act on before the bill arrives.
05The build prompt
Paste this into an AI coding tool (such as Claude, ChatGPT, Lovable or Replit) to build your own version.
Build me a canvas for running several coding agents at once, like Skribbl. Requirements: - Electron plus React, with React Flow for the pan/zoom canvas and xterm.js plus node-pty for the terminals. No web version, this is a desktop app. - Each node is a real shell in its own process, spawned with node-pty in a working directory I pick per node. Double click empty canvas to add one, and let me type in it exactly like a normal terminal. - Node position, size, title and cwd persist to a JSON file in the Electron userData folder, restored on launch. The shells themselves do not survive, say so in the UI rather than pretending they do. - Spawn each shell inside `tmux new-session -A -s <nodeId>` so a crashed window can reattach to the running agent instead of killing it. Budget real time for this, the reattach and scrollback restore is the fiddly part of the whole build. - Drag from one node to another to draw a link, and store the links with a direction. Right click a link to delete it. Links are just recorded relationships for now, no message passing between agents. - A top bar counting tokens per node, parsed from each agent CLI's own usage output. Keep input, output, cache write and cache read as four separate counters, never sum them into one number, a cache read costs about a tenth of an input token. - Out of scope: multiplayer, cloud sync, code signing, and any status indicator you would have to infer by scraping terminal output. Guessing wrong there is worse than showing nothing. - README with the node-pty rebuild step for your Electron version, which is the one thing that will break on a fresh machine.
06Open-source starting points
- claude-squad: open-source terminal manager for multiple agents in tmux and git worktrees
- Crystal: open-source desktop app running parallel agent sessions in worktrees
- vibe-kanban: open-source board for queueing and reviewing coding agent tasks
App prices, verdicts, alternatives and build prompts are adapted from Can I Vibecode It? (MIT License, © 2026 Rob Hallam). Each price shows the date it was checked and its source. Prices change; confirm on the vendor's site before you decide.
Get new verdicts in your inbox.
One short email when new verdicts land: what AI can now do for you, and what it still gets wrong. No spam. Unsubscribe anytime.