Is it safe to run Langflow with auto-login on a server?

No. With auto-login every visitor is a superuser who can run code and use your stored provider keys. Use it only for local experiments.

Where should I check the latest settings?

The Langflow documentation pages for authentication, environment variables, and security. Names and defaults change between releases, so confirm against the version you deploy.

Langflow Module 10: Deployment and Security Best Practices

Introduction

Langflow can run arbitrary code and call paid APIs with your credentials, so a careless deployment is a serious risk. This final module covers the settings and habits that make a deployment safe: authentication, secrets, network exposure, and restrictions for untrusted users.

Prerequisites

Modules 1 to 7
Basic familiarity with Docker and environment variables

Step-by-Step Guide

Step 1: Never expose Langflow directly

The documentation warns against exposing Langflow's ports directly to the internet without proper security. Put it behind a reverse proxy with HTTPS, restrict network access where you can, and treat it as a platform that executes code with no tenant isolation, so isolation is your responsibility.

Step 2: Turn on real authentication

Auto-login makes everyone a shared superuser, which is fine on a personal laptop and wrong for a server. Disable it, set a strong superuser password, and generate a secret key. Also disable open signups.

# .env
LANGFLOW_AUTO_LOGIN=False
LANGFLOW_SUPERUSER=admin
LANGFLOW_SUPERUSER_PASSWORD=use-a-long-random-password
LANGFLOW_SECRET_KEY=generated-below
LANGFLOW_NEW_USER_IS_ACTIVE=False
LANGFLOW_ENABLE_SIGNUP=False

Step 3: Generate a secret key and start with the env file

Create a random secret key and start Langflow with your environment file. Precedence is command-line options first, then the .env file, then system environment variables.

python3 -c "from secrets import token_urlsafe; print('LANGFLOW_SECRET_KEY=' + token_urlsafe(32))"

uv run langflow run --env-file .env

Step 4: Run it in Docker with persistence

For production, use pinned image tags rather than latest, persistent volumes, and a PostgreSQL database instead of the default local one. The repository's docker_example folder includes a Compose file with PostgreSQL. Upgrade by pulling the new image and restarting.

git clone https://github.com/langflow-ai/langflow.git
cd langflow/docker_example
docker compose up

Step 5: Harden for untrusted users and the network

Set CORS origins explicitly instead of the permissive default, use secure cookie settings behind HTTPS, and keep webhook authentication enabled. If people you do not fully trust can build flows, disable custom components, block code interpreter components, and restrict local file access. Enable SSRF protection so flows cannot be used to reach internal services. Set LANGFLOW_REMOVE_API_KEYS to strip keys from stored flows.

LANGFLOW_CORS_ORIGINS=https://your-app.example.com
LANGFLOW_ALLOW_CUSTOM_COMPONENTS=false
LANGFLOW_BLOCK_CODE_INTERPRETER_COMPONENTS=true
LANGFLOW_RESTRICT_LOCAL_FILE_ACCESS=true
LANGFLOW_SSRF_PROTECTION_ENABLED=true

FAQs

Is it safe to run Langflow with auto-login on a server?: No. With auto-login every visitor is a superuser who can run code and use your stored provider keys. Use it only for local experiments.
Where should I check the latest settings?: The Langflow documentation pages for authentication, environment variables, and security. Names and defaults change between releases, so confirm against the version you deploy.

Summary

Treat Langflow as code-execution infrastructure: keep it off the open internet, enable authentication with a strong password and secret key, persist data in PostgreSQL, and lock down custom components, files, and network access for untrusted users.