Setting this up for a coding agent instead? Follow Set up a coding agent. It installs the envlink skill, which walks the agent through the same loop.
1
Install envctl
envctl in /usr/local/bin, or ~/.local/bin when that isn’t writable. See Install for other ways.2
Log in
In the console, open Settings → CLI tokens and create a token. Then:The token is stored in
~/.envlink/credentials.json, readable only by you. On a CI runner, set ENVLINK_API and ENVLINK_TOKEN instead.3
Make this machine the sandbox
From the folder you want to work in:envd starts in the background and joins your organization’s tailnet. Share paths you use later are relative to this folder.
4
Lease a build box and share your folder
linux-build-3f2a, once it’s ready, usually in three to four seconds. Your folder syncs to its /work. Files your .gitignore leaves out stay home, and so do secret files like .env.5
Build there
/work/out is on your machine by the time exec returns:6
Release it
linux-build.Try a database
127.0.0.1:5432, carried over the tailnet. “Ready” means it accepts connections. See Services on localhost.
Next
Concepts
Environments, templates, shares, projects and snapshots.
File sync
What travels, what stays home, and how to change it.