100% static · no database

Shortlink

The whole destination is packed into the link itself — nothing is stored anywhere.

URL

Emoji mode

Password — optional, encrypts the destination

API

Returns the short link as plain text — the body is the link and nothing else. Failures are one error: … line.

GET /s/api/gen/#<url>
       ?emoji=1   allow the emoji alphabet

The url rides in the fragment, so it never reaches a server.

Preset links

968 well-known destinations ship with the site and cost one or two characters, because the code is a lookup rather than an encoding — no general codec reaches these lengths.

/s/#0  google.com    /s/#g   github.com
/s/#1  youtube.com   /s/#py  python.org
/s/#9  claude.ai     /s/#f   figma.com

Common sections compose onto a preset host with a dot, so those stay tiny too — 407 paths across every host, 400,488 combinations.

/s/#g.a       github.com/about
/s/#g.a.d     github.com/about/docs
/s/#bR.a.y.t  mozilla.org/about/legal/terms

Sections stack up to four deep — each part after the host is another section.

Keys come from the same 81-character url-safe set the codes use, so there is room for 6,642 of them.

Only the destination itself is preset. youtube.com/watch?v=… encodes in full as usual.

Password protection

Fill in the password field and the destination is encrypted with a key derived from it (PBKDF2 + AES-GCM). Only the ciphertext is in the link; the password is not, so it has to be typed on arrival. Costs about 47 extra characters.

curl -s https://austin-code.com/s/cli.js \
  | node - --password hunter2 "https://example.com"

Anyone with the link learns nothing about where it goes. A wrong password fails outright rather than decrypting to something plausible.

From a terminal

This site is static, so nothing computes a link server-side and curl alone cannot drive the endpoint above. Fetch the codecs and run them locally instead:

curl -s https://austin-code.com/s/cli.js \
  | node - "https://example.com/long/url"

  --emoji             allow the emoji alphabet
  --decode <code>     the other direction
  --password <pw>     lock or unlock
  --base <url>        another deployment

cli.js pulls codecs.js and runs the same four codecs this page does, so it always agrees with the site.