How to securely open up your server for RCE and "lambda functions"

I’ve previously written about how I exposed an endpoint for RCE, and how I could do it securely. Now I’ve implemented a component into the project, allowing you to dynamically declare such HTTP endpoints yourselves, using “WYSIWYG”.

Warning, you still need to have basic IT and security knowledge to be able to securely create such endpoints!

The point is that these endpoints accepts either code, natural language, or a combination. Then it executes the resulting code on the server, and returns the response back to caller.

Since everything can be capability constrained using RBAC, you can generate endpoints such as:

For then to have these guys’ agents freely generate whatever code they need, within the capability constraints declared in your policies. And optionally authenticate their existing MCP enabled harness, using OIDC, optionally on their own infra (with a little bit of effort). Think *“lambda function in Claude, Codex, etc, transparently given correct access to everything in anything the company has of private data”* - Allowing the AGENT to decide what CODE it actually needs, instead of deterministically create an *“API”* before hand …

Check out the following video where I’m illustrating the concept ==> [https://www.youtube.com/watch?v=pzdTfe2Sg_A](https://www.youtube.com/watch?v=pzdTfe2Sg_A)