Excited to share the multiplayer features we built to make running computer use agents easy. Here's a quick video of Claude ordering pizza and setting up Claude Code on the same computer.

we're focused on creating the foundation for humans and agents to do real work together on a computer. And we're building the core primitives to give a computer to your AI agent.

Our cloning feature is especially useful if you're already logged into your accounts and have apps installed for your agent. From your workspace view, simply press: - Tab - Cmd C + Cmd V You can spin up 10 clones of the same agent to parallelize tasks synchronously, such as form filling, outreach, or cron jobs.

You can access our new API references for templating, cloning, autostop, websockets endpoints, human-agent collaboration, and much more here:

You can use Orgo to give your agent a computer. Would love your feedback and excited too see what you build!

@vesper792 Should we cover this week chat 👀

@vesper792 thursday?

@vesper792 Check TG

the hard part of multiplayer computer use isn't the orchestration layer, it's that the OS only has one focus stack, one clipboard, one frontmost modal. two agents racing for the active window quietly corrupt each other's state and you don't see it until a paste lands in the wrong app.

this feature is useful for a human and agent to multitask or the human can use the computer and the agent can chime in when it needs to take over we give each agent an isolated computer. in most situations, it makes sense to give 1 agent per computer and have them connect to the same slack, g workspace, or acounts to work together we're not focused on trying to do 4 agents per computer we are instead doing 4 agents with computers to work together

the handoff is where this falls apart. agent uses the computer for 20 min, hands back, and i have no idea what state it left the browser in or which files got touched. isolated environments fix the safety part, not the resumability one.

we have a cloning feature that solves this the agent does work for 20 minutes, the context and results can be handed off or you can clone a computer which has exact same logins, browser sessions still logged in, all apps and context and data all the same

computer cloning for agent handoff is the right primitive. the gnarly part is state reconciliation when both the clone and origin run, most setups just throw one branch away instead of merging.

Cook

most founders miss this... agent automation really shines when you ditch technical oversight during execution, that's where time gets saved

which library you using for multiplayer?

$orgo on sale too. Get some coins while you can people

@austin_hurwitz thoughts?