Skip to main content

Connecting the CLI

Once scd-server is running, each developer connects their local scd CLI to it using a personal CLI token.

Team subscription required

scd-server requires a Team subscription. The scd CLI (Starter) works independently without scd-server. View plans →

Step 1 — Create a user account​

An admin opens the scd-server Admin UI and navigates to Admin → Users, then creates an account for the developer. The personal CLI token is shown once at creation — the admin copies it and shares it with the developer securely (for example, via a password manager or encrypted message).

If a token is lost, the admin can regenerate it from the same page. The old token is invalidated immediately.

Step 2 — Configure the CLI​

On the developer's machine:

scd configure --central-url http://your-server:3000
scd configure --token <personal-token-from-admin>
scd doctor

scd doctor confirms the connection, license tier, expiry, and AI provider status.

Step 3 — Activate the installation​

Run a scan to register the developer's machine with their account on the server:

scd scan

The first scan creates the installation binding. Until this step is completed, the developer's installation shows as Pending first scan in Admin → Users on the server.

Adjust timeouts if needed​

The defaults work for most setups:

scd configure --server-timeout 30s    # regular API calls (default: 30s)
scd configure --deep-timeout 20m      # deep analysis (default: 20m)

Deep analysis runs AI on your findings — 10–20 minutes is normal for large codebases.

Push local scan history​

If a developer has been scanning locally before connecting to scd-server, push their existing history:

scd sync --history

This is idempotent — safe to run multiple times.

After connecting​

From this point on, every scd scan pushes results to scd-server automatically. No extra commands needed.

The team dashboard at http://your-server:3000/dashboard shows aggregated findings across all connected developers.

Distributed teams (Tailscale)​

If your developers are not on the same network, Tailscale is the recommended way to make scd-server reachable:

  1. Install Tailscale on the machine running scd-server and on each developer machine
  2. Use the Tailscale IP or hostname as --central-url
  3. No firewall rules or port forwarding needed

Headscale is a self-hosted alternative to the Tailscale coordination server if you prefer to avoid any external dependency.