System Requirements
Hardware
Minimum is enough to run the application with 1-2 database connections under light load.
Recommended provides comfortable headroom for multiple connections and concurrent queries. The image runs the L0 regex prompt-injection guard out of the box; the L1 ML layer (DeBERTa ONNX, ~180 MB) is included but opt-in via
STUDIO_PROMPT_GUARD_ML_DOWNLOAD=true - see Security.- 4-8 CPU cores, 8 GB RAM, 20 GB disk
Software
Quick Start
1
Pull the image
2
Generate secrets
3
Run the container
4
Verify it started
http://localhost:3000.
First-Run Setup
On first visit, the app detects that no admin account exists and shows the Setup screen:- Create Admin Account - enter a username and password (minimum 8 characters). This is the only admin account for the installation.
- Login - after setup, you are automatically logged in and redirected to the Dashboard.
- (Optional) Enable 2FA - go to Settings, open the 2FA section, and scan the QR code with an authenticator app (Google Authenticator, Authy, 1Password, etc.). Save your recovery codes.
First Steps
1
Create a Connection
Go to the Connections page, click New Connection, select your database type, enter credentials, and test the connection.
2
Build a Tool
Go to the Tools page, click New Tool, select your connection, write a SQL query with parameters (e.g.,
SELECT * FROM users WHERE id = :user_id), and save.3
4
Deploy
Go to the Server page and click Deploy. See Deployment for details and MCP client configuration.
5
Connect a Client
Go to Server > Connect tab to get the MCP endpoint URL and configuration snippets for Claude Desktop, Cursor, or Claude Code.
Verify Installation
Openhttp://localhost:3000 in your browser - you should see the Setup screen (first run) or the Login page.
You can also verify via the terminal:
Next Steps
- Visual Flow Editor - drag-and-drop canvas for managing your entire configuration
- Configuration Reference - all environment variables and server settings
- Security Guide - encryption, authentication, and hardening
- API Reference - complete REST API documentation
- Agent MCP - AI agent access via MCP protocol
- Troubleshooting - common issues and solutions