Capy is a cloud IDE for orchestrating multiple AI coding agents in parallel, aimed at engineers who want to run entire sprints of work across isolated environments instead of babysitting a single assistant in their editor. Founded in 2024 by Nalin Semwal and Justin Sun, Capy is a Y Combinator company (Fall 2024) based in San Francisco , and the product has grown into a web-based development environment built from the ground up for multi-agent AI orchestration, allowing engineers to run up to 25 concurrent AI coding agents in parallel .
This launch introduces Capy Desktop, a native client that lets developers drive those cloud agents from their own machine. The desktop app is available for macOS, Windows, and Linux as a native app with tabs, deep links from Slack and the browser, and automatic updates . The pitch in the launch post is remote control plus access to local files and shells from any device, so a session started on a workstation can be picked up from a laptop, and work keeps running in the cloud after the lid closes.
The timing matters because the parallel-agent category is crowding fast, with CLI wrappers and worktree managers competing for the same workflow. Capy positions itself as a cloud IDE built from scratch for parallel development with isolated VMs and a two-agent model , and the desktop release pushes that orchestration layer out of the browser and into a daily driver surface alongside Slack, GitHub, and the terminal.
Comments (15)
Sign in to join the discussion.
N
Nadia Ravenko12d ago
Close your laptop and it keeps going in the cloud is a hell of a line, but what happens when my cat closes my laptop at 3am mid-refactor.
M
Maya Thornbury12d ago
curious what the actual close-laptop-to-cloud session completion rate looks like. the hero demo is one thing, p50 is another.
T
Theo A.12d ago
ok wait, the lid-close-to-cloud handoff is actually the first novel coding agent UX I've seen this quarter. Everyone else is just reskinning a chat box.
V
Vikram S.12d ago
the remote-handoff pattern reminds me a lot of what the Warp and Replit folks have been circling. capy might actually ship it first.
R
Rin Hasegawa12d ago
silly question but if I remote into my mac from my phone and the agent runs rm -rf, which device gets blamed in the logs?
M
Marco Peruzzi12d ago
the demo cuts feel suspiciously clean. staff eng brain asks: is this live or is this a very loved screen recording?
G
gm.sol12d ago
orchestrating agents across devices is literally a compute marketplace waiting to happen. $CAPY when.
H
Hana Okafor12d ago
how many people built this? if it's under 6 I'm going to go lie down.
D
Daria Milanovic12d ago
linux day-one is a nice flex. is there a CLI or just the desktop app, and where do the docs live?
P
Priya Balasubramanian12d ago
need to know the rate limits on remote shell invocations before I let this near my prod box. also webhooks for session events, pretty please.
F
Felix Brandt12d ago
this is the fourth coding agent launch I've seen this week and the first one that didn't open with a cursor comparison. refreshing.
Y
Yuki Tran12d ago
the thumbnail is doing a lot of heavy lifting here and I respect it. clean typography, no stock astronaut, we're healing.
Z
zeno12d ago
pretty sure this is the final coding tool that'll be marketed to humans. next year it's just agents buying licenses from other agents.
K
Kwame Boateng12d ago
the orchestration layer here sounds a lot like the multi-agent scheduling work out of CMU last year. is there a writeup on how tasks get routed?
L
Lena Fitzgerald12d ago
hot take: the tweet buries the lede. 'close your laptop and it keeps going' should've been the headline, not line three.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Capy
The best AI software engineer
The best AI software engineer
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Capy is a cloud IDE for orchestrating multiple AI coding agents in parallel, aimed at engineers who want to run entire sprints of work across isolated environments instead of babysitting a single assistant in their editor. Founded in 2024 by Nalin Semwal and Justin Sun, Capy is a Y Combinator company (Fall 2024) based in San Francisco , and the product has grown into a web-based development environment built from the ground up for multi-agent AI orchestration, allowing engineers to run up to 25 concurrent AI coding agents in parallel . This launch introduces Capy Desktop, a native client that lets developers drive those cloud agents from their own machine. The desktop app is available for macOS, Windows, and Linux as a native app with tabs, deep links from Slack and the browser, and automatic updates . The pitch in the launch post is remote control plus access to local files and shells from any device, so a session started on a workstation can be picked up from a laptop, and work keeps running in the cloud after the lid closes. The timing matters because the parallel-agent category is crowding fast, with CLI wrappers and worktree managers competing for the same workflow. Capy positions itself as a cloud IDE built from scratch for parallel development with isolated VMs and a two-agent model , and the desktop release pushes that orchestration layer out of the browser and into a daily driver surface alongside Slack, GitHub, and the terminal.
Comments (15)
Close your laptop and it keeps going in the cloud is a hell of a line, but what happens when my cat closes my laptop at 3am mid-refactor.
curious what the actual close-laptop-to-cloud session completion rate looks like. the hero demo is one thing, p50 is another.
ok wait, the lid-close-to-cloud handoff is actually the first novel coding agent UX I've seen this quarter. Everyone else is just reskinning a chat box.
the remote-handoff pattern reminds me a lot of what the Warp and Replit folks have been circling. capy might actually ship it first.
silly question but if I remote into my mac from my phone and the agent runs rm -rf, which device gets blamed in the logs?
the demo cuts feel suspiciously clean. staff eng brain asks: is this live or is this a very loved screen recording?
orchestrating agents across devices is literally a compute marketplace waiting to happen. $CAPY when.
how many people built this? if it's under 6 I'm going to go lie down.
linux day-one is a nice flex. is there a CLI or just the desktop app, and where do the docs live?
need to know the rate limits on remote shell invocations before I let this near my prod box. also webhooks for session events, pretty please.
this is the fourth coding agent launch I've seen this week and the first one that didn't open with a cursor comparison. refreshing.
the thumbnail is doing a lot of heavy lifting here and I respect it. clean typography, no stock astronaut, we're healing.
pretty sure this is the final coding tool that'll be marketed to humans. next year it's just agents buying licenses from other agents.
the orchestration layer here sounds a lot like the multi-agent scheduling work out of CMU last year. is there a writeup on how tasks get routed?
hot take: the tweet buries the lede. 'close your laptop and it keeps going' should've been the headline, not line three.