MCP server · Security

CSL-Core safety policy server

Your AI can write, check and test safety rules for other AI agents, so those rules can't be talked around.

CSL-Core is a tool for writing safety rules for AI agents and checking them before anything runs. The rules live in small text files, not inside the AI's instructions, so the AI cannot be tricked into breaking them. It is handy for anyone who lets an AI agent do real things, like move money or touch customer data.

What is an MCP server? The 30-second version

On its own, your AI can only chat. An MCP server is a small helper program that gives your AI a new skill or a connection to another tool. This one connects your AI to CSL-Core, a safety rule engine, so your AI can write a rule file, check it for mistakes, and test it against example situations. You stay in charge: the AI drafts and checks, you decide what to keep.

What this MCP server does

You describe the safety rule you want in plain English, and your AI turns it into a CSL policy file. It then sends that file to CSL-Core, which checks the logic for contradictions using a maths engine. If something clashes, you get told and can fix it. You can also test the policy against made-up situations, like a junior user trying to delete a record, and see whether it would be allowed or blocked. Finally, you can ask for a plain-English summary of any policy file.

Flow diagram: you ask your AI “Check my safety policy for contradictions”, on your own computer the CSL-Core safety policy server works with CSL-Core on your computer, and you get back pass, problems or ALLOWED / BLOCKED.

Click to zoom

What you can do with it

Turn a plain-English safety rule into a starter CSL policy file
Check a policy file for logical contradictions before you use it
Test a policy against example inputs and see ALLOWED or BLOCKED
Get a plain-English explanation of what an existing policy does
Catch rules that overlap or cancel each other out
Try out a rule change without touching your live setup

Try asking your AI

“Write me a safety policy that blocks transfers over 5000 dollars unless the user is an admin”
“Check this policy file for contradictions and tell me what is wrong”
“Test my policy with a junior user trying to delete a record”
“Explain in plain English what this CSL policy actually does”

What it gives back to you

You get answers in the chat, not files to hunt down. When a policy is checked, you get a pass or a list of problems in plain words. When you test a situation, you get ALLOWED or BLOCKED plus the name of the rule that decided it. When you ask for an explanation, you get a short summary of the rules in everyday language.

Before you start

What you need

Python installed on your computer
The csl-core package with the MCP extra, installed with pip
An AI assistant that supports MCP servers, like Claude Desktop, Cursor or VS Code

Good to know

These are safety rules for AI agents that can take real actions, so a wrong or missing rule can let something through; always test a policy before you rely on it.

Install it with your AI

Add CSL-Core safety policy server to your AI, no technical skills needed

You don't install anything by hand. You copy one prompt, paste it into an AI that can work on your computer, and it checks, installs and connects the server for you, asking you when it needs something.

Sign in to get the install prompt

Members get a ready-made prompt that lets the Claude desktop app check CSL-Core safety policy server, install it and connect it for them, step by step. You don't need any technical skills: you copy, paste and answer a few questions. Your connected AI can also find and install any of the 4,066 MCP servers here for you.