symflip is agentless: nothing runs on your server that it put there. Every deploy happens over an ordinary SSH connection — the new release is built beside the live one, switched in with a single atomic step, then health-checked. Here's exactly what happens.
main @ a1b9f301.2s12814sA deploy streaming live in symflip — every step, in your browser.
symflip opens an encrypted SSH connection to your server using a key you hold, pinning the host key on first contact so the identity can't be swapped later. There is no agent to install, patch, or babysit — nothing extra runs on the box.
It fetches the exact commit you're deploying into a fresh, timestamped
releases/ directory. The release that's currently serving is
never touched.
main @ a1b9f301.2sreleases/1280.3sYour build, migration and pre-activate steps run inside the new release while the old one keeps serving every request.
If any of them fails, the switch simply never happens. The live release is untouched and the deploy stops safe. You never ship a half-built release.
Going live is a single atomic step: the current symlink is
re-pointed at the new release with one rename, then the app
is reloaded to pick it up. There is no moment where the
server points at a half-copied directory.
Traffic is served by the old release right up until the instant it's served by the new one — which is why a release costs zero downtime.
releases/ 126/ ← older, still kept 127/ ← previous, kept warm 128/ ← new, fully built current → releases/128 (atomic swap)
symflip keeps old releases on the box, so rolling back is another instant symlink flip — no rebuild, no scramble. One click returns you to the last good version.
And if a release goes live but then fails its health check, symflip can roll itself back automatically to the previous release. Turn on auto-rollback and a bad deploy heals itself before you even reach for the button.
Add more servers to a project and they ship in lock-step: every server builds the new release first, and none goes live until all are ready. Then each switches over, and if one fails to switch or the release fails its health check with auto-rollback on, the ones that switched roll back. No half-built fleet.
A fleet deploy in symflip: every server builds before any of them switches.
Connect a repo and a Linux box, click deploy, and watch the whole thing happen live. Your code and servers stay yours — symflip just makes shipping effortless.