features

create a vm, run commands, and destroy it when you are done

choose a size or set cpu, memory and disk yourself

boot with rust, node, python, or a desktop and browser

build from a spec or upload an image, then pin a version

save a vm's files and restore them into a new vm

memory and files, in one snapshot

a vm sleeps on its own disk and wakes in 0.1 to 0.3 s when next used

fork one state into many runs

set a lifetime and idle timeout, or keep a vm running without either

wait for room in a region, or fail at once when it is full

upload your checkout, run a command, fetch its output files, and end the vm

try opencode in a vm from your browser

claude code or codex into a vm with one command, back with another

watch an agent, check its state, and pull its commits and conversation back

agents drive real desktops: screenshots, mouse, keys

watch any of them live, or take over

share a screen or terminal through a view only link you can revoke

record a broadcast into your bucket and play it back later

write to a vm's scratch disk and grow it while the vm runs

create private buckets, upload files, and download or delete objects

grant temporary access to one object for uploads or downloads

reuse an image when its spec, files and base match

expose a vm's http port at a named url on kyora.run

serve a public url on your own domain

allow package registries, named hosts, the internet, or no network

a named region is strict, global picks the nearest with room

each job gets a fresh vm that ends when the job finishes

choose a runner's size, image, region and network rules

choose repositories, cap concurrent jobs, and inspect runs in the dashboard

several coding agents review the same branch or pull request independently

merge matching findings and ask another reviewer to challenge single findings

review github pull requests with kyora models

ask model families through mcp, with answers now or in the background

assign work to another model family, with optional file edits

save and load claude code, codex and opencode go or zen logins by name

read remaining quota and counted tokens across stored accounts

vm usage is priced by allocated cpu and memory seconds, plus written disk

inspect monthly costs by vm, region, storage and day

share vms, snapshots and tokens under one account with member roles

create scoped tokens with an expiry and revoke them when needed

open a ticket and read or reply from the cli or dashboard

stream command output, transfer files, and use json output or shell completions

detach from a terminal and attach again while its command keeps running

inspect cpu, memory, disk and network usage from the cli

create vms, run commands, manage snapshots and share screens over http