I will review and secure your mcp server or ai agent setup

An MCP server is a door into your data. Most of them get built in a hurry, with a token that can do everything and tools nobody restricted.

I look at that door the way an attacker would, then close it.

What I check:

• What the server actually exposes - every tool, every argument, what it can reach
• Tokens and keys: where they sit, what scope they have, who can read them
• Whether an agent can be talked into calling something it should not
• Logging - would you even know if something went wrong last Tuesday?

What you get:

• A findings list in plain English, worst first, each one with the fix
• On Standard and up, the fixes done and re-checked, not just described
• A short note on what to watch after I am gone

I have run my own servers behind Cloudflare since 2023 and I build security monitoring for a living (secmon.io). Same work, pointed at your agent stack.

Send me the repo or the config. A read-only copy is fine.

Not included: writing a new server from scratch (that is my other gig), pentesting your whole infrastructure, compliance paperwork.