Firas KhalifeSafe MySQL & AWS MCP servers
I built two internal MCP servers so that AI tools can explore our MySQL databases and our AWS setup safely: read-only, isolated, and redacted for secrets and personal data.
BeforeUnofficial servers or personal credentials: one slip from a write, with secrets and personal data exposed to the AI tool.
AfterIsolated, read-only, secrets and personal data redacted.
How they keep it safe
1 / 5Before: one slip from a write, and data exposedPeople were reaching for unofficial MCP servers, or pointing an agent at a database with their own credentials, where one bad prompt can become a write, and secrets, sensitive data and personal information land straight in the AI tool.
MCP clientAI tool
beforeOwn credentials
MCP serverMySQL
MCP serverAWS
Database
Database
read onlyAWS account
- 1Before: one slip from a write, and data exposed
- 2Its own read-only user, or role
- 3A container per database
- 4Redacted on the way back
- 5Zero setup
Where they’re used
01Safe, AI-assisted exploration of our databases and AWS
02 · Release merge dry-runReplays each release’s migrations against the staging schema and data through the read-only MySQL server →03 · QA × AI seriesTwo sessions: “MCP servers” and “MCP servers in action” →Why I built them
I built these servers to remove that: each runs under its own read-only user or role, each database gets its own container, and responses are redacted for secrets and personal data for compliance. They work out of the box.
What each server guarantees
- Identity
- Its own read-only database user, or AWS role
- Isolation
- One container per database
- Data
- Secrets and personal data redacted
- Setup
- None: works out of the box