Worked on getting sandboxes into T3 Code Instead of running on main or a new worktree, it spins up a fresh VM, run the agent there, and adds it as a remote environment on all devices No more managing memory on my machines or balancing worktrees

Works by having your T3 server hold a Boat key and owning the lifecycle. Creates the VM, installs T3 inside it, and every client you've paired picks the sandbox up automatically. It's just another T3 environment. Still lots and lots of work to make this good, but it's a cool prototype

would love to have this upstream! My machines keep oom killing stuff because I have too many agents working at the same time. T3 is helping me spin up more ideas than I have compute for

I don't understand the use case of remote agent. your repo copy is in local then you pushed it in remote machine, agent works there. you only saved local resources.and you probably pushed claude credentials too. if you just want to isolate the agent, local docker makes more sense. it's just unnecessary complexity with diminishing returns.

If you have an Anthropic or OpenAI subscription, you can use their cloud sessions with subs, and they're generous with the specs, for "free"

bruhh been building exactly this in @wsplabs for the past 2 weeks... We literally ran 600 threads in boat using wsp to build wsp itself 💀

thanks for doing this!

you're the goat sir

@theo we need a plugin system for T3 Code similar to OpenCode v2 to let us bring in features like this!

wow! love this man

YES!

fresh vm per task is clean but you pay setup every time, snapshotting a warm one and forking it per agent gets the same isolation without rebuilding the world

Does the VM stick around when the agent finishes, or do you pull the changes back and shut it down?

This is pretty cool

Can a sandbox be created with more than one repository?

fresh vm as the remote env is what collapses the worktree juggling. my laptop still pays the agent memory bill

I’d love to try this!

Dope

This is very cool and I'd love to try a workflow like this

Does the same VM survive a T3 server restart, or does it get recreated?

can you also move the connections to the under the compose chat? why would i need to go top right corner to see which device I'm using?

Count me in

but we dont have to login inside sandbox?

That's sick! just set up my own VM on my homelab to do fully remote dev - would be interesting to see how i can selfhost something like this!

@conchcli are supported and Daytona. through CLI, create sandbox and run Claude, Codex any ai agents.

Fresh remote sandboxes sound like a much cleaner way to manage agent workflows

This is exactly what I need 👏👏

Remote VMs beat local worktree juggling every time

@maria_rcks would be great to have VM support on T3 Code: boat, daytona, ...

@bil0090 This would be a great addition to T3 Code!

this is the missing piece

Looks awesome would love this get merged!

cool, hope this gets merged

Please add it in T3 code

@jullerino we need this in T3 Code!

@theo this would be game changer!