envden keeps the .env files in your repos the same on every machine you work on. It syncs whole files, byte for byte, so comments, order and blank lines survive. Everything is encrypted on your devices before it leaves them, and it syncs through a small server you run yourself on your Tailscale network.
Alpha. Commands and file formats may still change.
Docs: envden.theclau.de/docs
# macOS and Linux
curl -fsSL https://envden.theclau.de/install.sh | sh
# Windows (PowerShell)
irm https://envden.theclau.de/install.ps1 | iex
# Homebrew
brew install notsuhas/tap/envden
# npm
npm i -g envden
# Scoop
scoop bucket add notsuhas https://github.com/notsuhas/scoop-bucket
scoop install notsuhas/envdenThe server is the same binary (envden serve), or the image ghcr.io/notsuhas/envden:0. Binaries and checksums are on the releases page.
On a home server (Docker):
docker run -d --name envden --restart unless-stopped \
--read-only --cap-drop ALL --security-opt no-new-privileges:true \
--tmpfs /tmp:uid=10001,gid=10001,mode=1777 \
--tmpfs /var/run/tailscale:uid=10001,gid=10001,mode=0700 \
-e TS_AUTHKEY="tskey-auth-..." \
-v envden-data:/data \
-v tailscale-state:/var/lib/tailscale \
ghcr.io/notsuhas/envden:0
docker logs envden # prints a one-time setup tokenReplace tskey-auth-... with a Tailscale auth key. The server joins your tailnet as envden, at https://envden.<tailnet>.ts.net.
On your first laptop:
envden setup --server https://envden.<tailnet>.ts.net # paste the setup token, choose a master password, print the recovery kitNo home server? envden setup --host-server runs the server on this laptop instead, with no token to paste.
On every other device:
envden device add # on a device that's already set up: shows a pairing code
envden join <server> # on the new device: type the codeDay to day:
envden status # what's out of sync (key names only, never values)
envden pull # get changes from your other devices
envden push # send your changes
envden ui # the same in a local browser dashboardEvery command and flag is in the CLI reference.
- Your env files live in
~/.envden/store, and each checkout links to its file, so every clone and worktree shares one copy. - Nothing syncs in the background.
pushandpullmerge edits per key and leave real conflicts for you. - The server stores only ciphertext and opaque IDs. A master password and a recovery kit protect the vault.
envden agent installlets coding agents check status while you keep the passwords and approvals.
The server can't read your files or add devices on its own. It can refuse service or ignore unlock expiry, and a lost device keeps whatever it already pulled. See Security for the model, DESIGN.md for the threat model and PROTOCOL.md for the cryptography. To report a vulnerability, see SECURITY.md.
For contributors. Needs Bun 1.4.
bun install
bun run --cwd apps/cli build # writes apps/cli/dist/envdenThe server image is built from apps/server/Dockerfile, and releases are described in docs/releasing.md.
| Path | Contents |
|---|---|
apps/cli |
The envden command |
apps/server |
The sync server (Hono on SQLite) |
apps/web |
Website and docs (Next.js, Fumadocs) |
packages/core |
Protocol, crypto, chain validation, merge, adapter interfaces |
packages/adapters/<kind>/<impl> |
Git, file placement, keychain, shell, Tailscale, SQLite |
packages/testing |
Contract tests and fakes for adapters |
docs/ |
Design and decisions, protocol spec, terms, adapter guide, release runbook |
Read AGENTS.md first; its rules apply to people and agents alike and lint enforces package boundaries, file layout, naming and comment limits.
bun run check # format, lint and type check (Vite+)
bun test
bun run fallow # unused code, cycles, duplication, complexityAfter changing a command's help text, run bun run docs:cli to regenerate the CLI reference; a test fails while it's stale.
Adding an adapter: see docs/adapters.md or run bun run new-adapter <kind> <impl>.