Quickstart

Install the CLI, authenticate through your browser, and get your first findings on the dashboard.

You need an invitation

SecureAI is invite-only. Your organisation admin invites you by email; accepting the invitation creates your account. If nobody has invited you yet, ask your admin to add you from Users.

Install the CLI

One command on macOS or Linux. You do not need an account to download it. Windows and manual downloads are on the Installation page.

Terminal

curl -fsSL https://get.corelayersecurity.com/install.sh | sh
secureai --version

Point the CLI at your control plane

Set SECUREAI_SERVER once in your shell profile so you never have to pass --server.

Terminal

export SECUREAI_SERVER=https://api.corelayersecurity.com

Log in

secureai login opens your browser, you sign in with the same account you use for the dashboard, and the CLI stores the resulting session locally.

Terminal

$ secureai login
  Opening your browser to sign in…
  Waiting for the callback on http://127.0.0.1:54123/callback

  Signed in as you@yourcompany.com (developer)

Confirm at any time with secureai whoami. Details of the flow are on Authentication.

Run your first scan

Point scan at the directory holding your prompts and tool definitions. It reads files statically. It never calls your model, so it costs nothing and is safe to run anywhere.

Terminal

secureai scan ./prompts

Findings are grouped by severity, each mapped to an OWASP LLM Top 10 category and a rule ID you can tune or disable.

Generate a report

Add --report to write a self-contained HTML assessment you can open in a browser or print to PDF.

Terminal

secureai scan ./prompts --report assessment.html

See it in the dashboard

Open app.corelayersecurity.com. Every run uploads to your organisation automatically, so your scan is already there under Scans, with findings under Findings.

Where to go next

Testing a live endpoint? TEST sends adversarial payloads at a running model and measures how many succeed.

Wiring this into CI? CI integration covers exit codes, machine-readable output, and failing a build on severity.

Want to tune the rules? Configuration explains secureai.config.yaml, severity overrides, and gates.