Video + Guide

Secure AI + Database Integration in 5 Minutes with DreamFactory MCP and Claude

Generate a read-only API for your database, auto-build a dashboard, and chat with your data—without exposing raw SQL or credentials.

Watch the 5-minute demo

Download DXT

npx @Dreamfactory/Create-Df-Mcp

AI + Databases, Done Right

Most “AI + data” demos look impressive—but often give large language models direct SQL access. That’s fast for a prototype and risky for an enterprise. The workflow in this video shows a different pattern: let AI orchestrate secure APIs, not write SQL.

• Fast setup: start with a simple .env containing DB credentials.
• Secure by design: DreamFactory MCP generates read-only, parameterized endpoints with RBAC and audit logs.
• Conversational access: use Claude Desktop via MCP to explore tables, schemas, and insights—no raw SQL exposed.

Step-by-Step: From .env to Dashboard

Try it now: df-mcp-v1.0.1.dxt or run npx @dreamfactory/create-df-mcp.

Why the MCP Gateway Matters

• Zero-trust access: treat AI like an untrusted client; all actions must be authenticated, authorized, validated, and logged.
• Parameterization & validation: neutralize prompt→SQL injection by enforcing schemas, types, ranges, and allowlists.
• RBAC & masking: scope access by role; hide PII with field/row-level policies before data reaches the model.
• No credential exposure: secrets stay server-side; the LLM never sees connection strings.
• Audit by default: immutable logs show who accessed what, when, and from where—SIEM-ready.

Deployment: Minutes, Not Months

Choose the fastest path for your environment:

• DXT rapid deploy: download the DXT and bring up the MCP server with sensible defaults.
• One-command setup: npx @dreamfactory/create-df-mcp for an instant project scaffold.

Databases supported out of the box: SQL Server, MySQL, PostgreSQL and 20+ more. Standard OpenAPI makes it easy to plug into RAG frameworks and agents.

FAQ

 

What exactly happens in the 5-minute demo?

 

Claude Code generates a read-only API and a dashboard from your .env, while Claude Desktop connects via MCP so you can chat with your data safely—no raw SQL or exposed secrets.

 

 

 

Why not let the LLM write SQL?

 

Direct SQL invites prompt→SQL injection, credential leaks, policy bypass, and performance regressions. An API gateway enforces parameterization, RBAC, masking, and logging.

 

 

 

Which databases can I use?

 

SQL Server, MySQL, and PostgreSQL are supported with consistent security controls and auto-generated OpenAPI docs.

 

 

 

Do models ever see credentials or schemas?

 

No. Credentials remain server-side in DreamFactory. Models only see vetted endpoints and structured JSON responses.

 

 

Can I integrate this with RAG or agents?

 

Yes. The OpenAPI spec and deterministic JSON responses make these endpoints ideal for RAG retrieval and agent tool

functions.

 

What about compliance (GDPR/HIPAA/SOC2)?

 

Zero-trust, RBAC, input validation, masking, data residency controls, and immutable audit logs provide the evidence trail auditors expect.

 

 

How do I try it?

Download the DXT package or run npx @dreamfactory/create-df-mcp for instant setup.

Conclusion

You don’t need AI to write SQL to get value from your data. In five minutes, you can generate secure APIs, spin up a dashboard, and talk to your database through a governed gateway. Stop exposing raw SQL—start orchestrating secure APIs.

Download DXT

npx @dreamfactory/create-df-mcp