All your agents.
All your projects.
All your machines.
Claude Code, Codex & more run on your own computers, on the subscriptions you already pay — many at once, in the background, always on. Leave the laptop at home and steer them over the internet from your phone, tablet, or desktop.
Agents supported: Claude Code, Codex, Cursor & more
revenue-by-customer-2025-05–2026-04.csv — 47 customers across 12 months, total recognised revenue $1,418,302.14. Top 10 are 60.9% of it.scripts/revenue-report-monthly.ts, scheduled it for the 5th at 09:00 UTC, and emailed you the first digest with both CSVs. Anything else?

Your experience.
Your agents' experience.
For you
Works from anywherenew
Leave the laptop at home: your agents keep working there, and you steer them over the internet from your phone. No VPN, no port forwarding, no same Wi-Fi.
Many agents, one chat list
Claude Code, Codex, Cursor, Gemini, Hermes, and Pi, side by side. Pick the agent when you start a chat; they all live in the same list.
Search everything
Full-text search across thousands of chats and message bodies — find what an agent said weeks ago.
Attach Files
Attach screenshots, PDFs, anything — drag them in, pick from the library, take a photo, or just paste from the clipboard. Attachments are end-to-end encrypted too.
Pinged when done
Push notifications when an agent finishes. Walk away from your laptop without losing the thread.
Queue the next message
Line up the next message while an agent is still working — it sends the moment the agent is free. Or pick a time and it fires on its own.
Background chatsnew
Move long-running chats out of your list. They keep working, stay out of the way, and don't ping you.
Mac, server, shared
Connect your laptop, desktop, or home server, and pick the right one for each chat. Share machines and use them together with colleagues and friends.
Import your chats
Add your machine and your existing sessions are imported as chats automatically. Keep working on the conversations your agents already have.
Parallel worktrees
Spin off any chat into its own git worktree so agents don't step on each other's branches. Flip a toggle when starting a new chat — the spawner sets it up automatically.
End-to-end encrypted
Message bodies are encrypted on your device. Our server stores ciphertext, never plaintext.
Check what's installed
The spawner — the only Zucchini piece that touches your computer — is on GitHub, so you can read the source before you install.
For agents
Agents that never quitnew
Turn on always-on and a Claude Code or Codex agent keeps running for months: it comes back after a reboot, compacts its context when it fills up, and keeps the standing instructions you gave it — and can update them itself.
Wakes on eventsnew
An agent can watch anything — a CI run, a log, an inbox, a price — and wakes up the moment something happens.
Its own schedule
Agents book their own follow-ups: check the deploy at 9, retry the job in an hour, send the weekly report on Friday.
A public linknew
An agent can give anything it runs on your machine an HTTPS link — for webhooks, sign-in callbacks, or a preview you open on your phone. Requests that arrive while the machine is offline wait and are replayed.
Dynamic Model Context
Agents prune the tool output they're done with and keep going on a leaner context — so a long chat doesn't grind to a halt. You see exactly what was freed, right in the timeline.
Questions
What is Zucchini?
A messenger app for talking to your AI agents from your phone, tablet, or computer. Every chat spawns a fresh agent in a project folder on your machine, runs the task, and streams the reply back into the chat.
Do I need to install anything on my computer?
Yes — a small helper called the Zucchini spawner. One-line install on macOS, Linux, or Windows. The spawner is what runs your agents locally; the app is the chat surface on top of it.
How much does it cost?
There's a free tier — one running agent at a time, no time limit, no card required. Zucchini Pro is $9.99/month for unlimited concurrent agents and unlocks on all your devices the moment you subscribe on any of them. Your agent usage — Claude, Codex, and the rest — is billed by each provider through your own account.
Do I need a VPS to run agents 24/7?
No. Your agents run on the computer you already have — a Mac, a Linux box, or a Windows PC — on the Claude or ChatGPT subscription you already pay. Turn on always-on for a chat and its agent keeps running and comes back by itself after a reboot. The computer has to stay on; a home server or an old laptop works well.
Can I share a machine with my team?
Yes — it's what Zucchini is built around. Set up a machine once, then invite other people to it and they get working agents on it with nothing to install: no Mac, no setup, no terminal. They just open a chat, and the agents run on your machine.
Is my data secure?
Message bodies are encrypted on your device with XChaCha20-Poly1305 before they leave it. The encryption key lives in your device's secure key store — the iOS Keychain on iPhone, iPad, and Mac, the encrypted keystore on Android — and never leaves your devices. Our server stores only ciphertext and never sees plaintext.
Do you store my code?
No. Your agents run on your computer, against your local files. Only the chat messages — encrypted — are synced through our server, so the conversation appears on all your devices.
What devices does Zucchini run on?
iPhone, iPad, and Apple Silicon Mac: App Store. Android: Google Play. Windows, Linux (x86_64), and Mac (macOS 26 or later): direct download from this site.
Can I bring my existing chat history?
Yes — automatically. When you add a machine, your local agent sessions are imported as chats so you can keep working on the conversations you already have. Claude Code history is imported today, with more agents being added. No migration, no copy-paste.
Does it work with Codex, Gemini, or other agents?
Yes, today. Zucchini works with Claude Code, Codex, Cursor, Gemini, Hermes, and Pi — all from the same chat list. You pick which agent to use when you start a chat.
Zucchini