Langflow’s auto-login signs in all users as the configured superuser by default, a behavior that depends on installation method (True for pip, False in official Docker images).

There are three defaults to consider changing: tighten CORS with LANGFLOW_CORS_ORIGINS and LANGFLOW_CORS_ALLOW_CREDENTIALS, avoid altering LANGFLOW_SECRET_KEY on existing installs, and ensure secure port binding and HTTPS by enabling LANGFLOW_ACCESS_SECURE and LANGFLOW_REFRESH_SECURE.

Release notes show version differences (1.10.3 vs 1.12.5) and advise users to verify the version in use; a broader hardening checklist and community resources are included for self-hosted AI security comparisons.

The /api/v1/auto_login endpoint returns a superuser token when auto-login is enabled and a 403 when it's disabled; testing from another machine can reveal open admin access.

A known security advisory GHSA-w794-rj3p-xv45 is cited regarding OS command execution when signed in, and auto-login expands that risk to any user.

Recommended security measures include disabling unrestricted superuser creation (LANGFLOW_ENABLE_SUPERUSER_CLI=False), exporting flows before changes, and using strong random passwords for the superuser.

To turn off auto-login, set LANGFLOW_AUTO_LOGIN=False, configure a specific superuser and password via environment variables, and start with an .env file.

For Docker deployments, avoid exposing port 7860 on all interfaces; bind to localhost or use tunnels/proxies, and pursue secure configurations for production.

On servers or shared machines, auto-login effectively makes any reachable user an admin, potentially enabling code execution on the host.