An agent that can fully modify itself. It ships as a single chat box. The UI, the server, its tools, its own system prompt — all of it is its own source code, and it can rewrite any of it. Ask for something and it changes itself, rather than handing you a separate artifact.

"Turn yourself into a Windows XP desktop." Real recording, from the factory state, middle fast-forwarded:

- Everything stays local. The agent, its code and your data live on your machine. No account, no cloud.

- Bring your own model. Any endpoint compatible with the OpenAI Responses API: URL, API key, model name.

- Breaking it is safe. A snapshot is taken before every boot; if the new code doesn't come up, it rolls back automatically. There's also a separate recovery window.

Real recordings, nothing edited by hand. The middle of each clip is fast-forwarded; the moment it reloads is real time. Above: it rewrote its React and CSS, rebuilt, reloaded — the chat became a window on the XP desktop, the panel it had grown earlier became another one.

"Become a crypto price desk, with real data." It couldn't read the web at all. It wrote itself a tool that pulls OKX quotes, added a scheduler that refreshes every minute, and built price alerts that fire a system notification.

"From now on you're the dungeon master." It rewrote its own system prompt. The UI became a text adventure with option buttons, and every move rolls real dice.

"Put throw new Error() at the top of your own main.js and restart." It crashed. The launcher rolled it back to the last good snapshot in seconds; it came back, wrote an incident report, then noticed the rollback had gone one step too far and recovered that work from the branch itself.

Open it and type into the prompt:

Add a to-do list on the right

Turn yourself into a vintage instrument panel

Give yourself a tool to read web pages

It edits its own code, reloads, and tells you what's new and how to use it. It answers in whatever language you write in.

macOS (Apple Silicon): download the dmg from Releases. On first launch it asks, in the chat, for your model settings: endpoint URL, API key, model name, context window size.

The API key is stored only in a local config.json (mode 600) and sent only to the endpoint you entered. A stronger model does noticeably better; everything above ran on DeepSeek V4 Flash.

launcher/ The only part of the install that never changes. Finds the workspace → seeds it on first run /

upgrades it → bookkeeping → loads the app.

Snapshots before every boot; two failed boots of the same version → roll back to the last good

one, the broken version is kept on a crashed-* branch.

The `iimos` CLI: status / snapshot / log / diff / rollback / reload / restart / add

Recovery window: loads none of the app's code; rollback, factory reset, a minimal rescue agent.

app/ The app as shipped (the seed). Copied into <data dir>/workspace/ (a git repo) on first launch;

everything the agent changes happens there. Ships as nothing but a chat:

desktop/ starts the local server → opens the window → reports ready to the launcher

server/ local server, 127.0.0.1 only, token-protected; conversation, event stream

ai/ model calls (Responses streaming, automatic retry)

agent/ the agent loop; tools shell / read / write / edit; context auto-compacted at 70%;

prompt.md is its base system prompt

ui/ React + TSX; after a change, `iimos reload` compiles with the esbuild the launcher ships

AGENT.md notes the agent keeps for its future self

The only contract the app has to honor (see launcher/CONTRACT.md): report ready to the launcher after starting, and keep user data in $IIMOS_DATA, not in the workspace. Everything else is fair game.

Data directory (macOS): ~/Library/Application Support/ai.iimos.desktop/

— workspace/ (app code, a git repo) · data/ (conversation, model config; never rolled back) · launcher/ (ledger, logs).

Recovery window: ⌘⌥⇧R (Windows/Linux: Ctrl+Alt+Shift+R), or the --recovery flag.

Node 22.5+.

npm install

npm run appUse a throwaway data directory:

IIMOS_HOME=/tmp/iimos-dev npm run appnpm test # launcher seeding / rollback / upgrade + app end-to-end (fake model)

npm run build:app # unsigned build

npm run ship:mac # sign + notarize + verify