Grass 2.0 - The always-on computer for your coding agents

Grass gives your coding agents their own always-on cloud computer. Run Claude Code, Codex, or OpenCode on Grass, then monitor progress, approve decisions, and push changes from your iPhone. Version 2 is faster, redesigned, and now live on the App Store.

Add a comment

Replies

Best
Hey Product Hunt 👋 Three months ago, we launched Grass here as a scrappy alpha with a simple idea: coding agents need a home of their own. Claude Code, Codex, and OpenCode are getting better fast. But running them on your laptop still feels fragile. You keep a terminal open. Your machine goes to sleep. You step away and lose visibility. Your agent needs a decision, but you are not there to answer. Grass gives your coding agents an always-on cloud computer, and gives you an iPhone app to stay in touch with them. Today we’re launching Grass 2.0, and it’s a much sharper version of the original idea. What’s new: → Grass is now on the App Store No TestFlight invite. No PWA workaround. Download the app and start running agents. → A full redesign We rebuilt the app from the ground up. Cleaner UI, faster flows, better session views, and a new icon. → Faster sessions Connecting to agents feels quicker. The app is more responsive. The whole experience feels more reliable. → Support for more agents Grass works with Claude Code, Codex, and OpenCode today, with more coming. → Built with real usage Developers are already running agents on Grass, and V2 is shaped heavily by what we learned from them. The core idea hasn’t changed: coding agents are no longer tools you babysit. They are workers you stay in touch with. Grass is their computer. Your phone is the control room. We use a BYOK architecture, and every new user gets 10 free hours with no credit card required. We’d love your feedback; especially on what agents you want us to support next, and what your ideal mobile workflow looks like. We’ll be in the comments all day 💚

 Congrats on the App Store release! Quick question: for a busy developer who wants to trust an agent to run longer jobs or respond to prompts while they’re away, what safeguards and visibility features does Grass provide to prevent unintended actions, surface important decisions, and let me audit what an agent did after the fact?

   Grass let's you pick permission modes which means you can have the agent ping you via notifications for more sensitive actions. Would love your feedback on the notifications experience :)

 The always-on angle solves the real pain, agents that stall because the local machine sleeps or loses context mid-session. Curious whether you handle multiple concurrent agents on the same VM or is it one agent per instance?

Honest question: what's the practical difference between Grass and just spinning up a cloud VM (EC2, Railway, Fly.io) and SSH-ing in with tmux? For a developer who already has that workflow, the setup cost is minimal and the VM stays running indefinitely.

The iPhone monitoring angle is interesting, but "approve decisions from your phone" is doing a lot of work in the pitch - how granular is that actually? Is it full terminal visibility + input, or more of a notification layer where you review outputs periodically?

Genuinely curious what V2 added beyond V1 that makes this worth switching to vs the DIY approach. The packaging as a dedicated product might be worth it for non-developers, but the target audience (people running Claude Code / Codex) are usually pretty comfortable with cloud infra.

 Fair question, Gal.

If you already have an EC2/Fly/Railway VM + SSH + tmux setup that you like, that absolutely works. Grass isn’t claiming the DIY path is impossible.

The difference is mostly effort and experience: Grass packages the VM, agent setup, mobile access, session visibility, and steering into one product instead of making you stitch it together yourself.

On approvals: today it’s closer to full session visibility + input from your phone.

V2 added the App Store app, a full redesign, faster sessions, Claude Code / Codex / OpenCode support, and a more solid end-to-end experience shaped by real usage.

Cool idea any specific real world example for it?

 we're excited about steering our coding agents from the app.

Seen this demo'd, awaiting arrival on Android! Excited for this one!

 Thanks Siddhesh :) Look forward to releasing the Android soon.

Giving coding agents an always-on home makes sense. BYOK keys and repo credentials living on an always-on cloud VM, with an agent running unattended, is a real surface. Is each session sandboxed, and do the keys persist on the VM between runs or get cleared when the session ends?

Does Grass handle partial approvals on risky commands, or is it all-or-nothing when you're away from your machine?

Grass 2.0 seems a lot greener than Grass 1.0, but that tracks with the idiom.