MCP server · Security
Toolwrit MCP server
by emektor
Put a written rulebook in front of your AI's tools, with spending caps and a tamper-proof record of every action.
Toolwrit is a small helper that sits between your AI and the tools it uses. It checks every single tool call against a rule file you write, blocks anything you did not allow, and keeps a record you can check later. It is handy for anyone who lets an AI touch real files, real money, or real customer data and wants proof of what happened.
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 an app or service. Toolwrit is a helper that sits in front of another MCP server, for example a file server, and acts like a strict doorman. Your AI still asks for things as usual, but every request passes through Toolwrit first, and only the ones your rule file allows get through.
What this MCP server does
You write a simple rule file that says which tools are allowed and under what conditions. When your AI asks to use a tool, Toolwrit looks at the request and checks it against your rules. If the request breaks a rule, Toolwrit refuses it and tells your AI exactly which part was wrong, so it can try a different approach. If the request is fine, it goes through to the real tool and the result comes back to your AI. Every decision, allowed or refused, is written into a log file that is chained together so nobody can quietly change it later.
What you can do with it
- Allow only specific tools and block everything else by default
- Set limits on how many calls, how much money, how many bytes, or how many seconds a run may use
- Block file paths that contain secrets like .env or id_rsa
- Check a rule file before a run to see what it would allow or refuse
- Get a plain-English summary of a rule file for a reviewer
- Verify that a log file has not been altered since it was written
- Compare one run against earlier runs to spot unusual behaviour
Try asking your AI
- “Can you read the notes file in my workspace folder?”
- “Write a summary of the CRM export into /srv/workspace/summary.md”
- “How much of my budget has this run used so far?”
- “Show me the record of every tool call from last night's run”
What it gives back to you
You get the normal result of whatever tool your AI asked for, such as file contents or a confirmation. When something is refused, you get a short message saying which tool was blocked and why, for example that a path was outside the allowed folder. You can also ask for a summary of a run showing how many calls were made, how much was spent, and a code that proves the log has not been changed.
Before you start
What you need
- Node.js 20 or newer, or Python 3.10 or newer
- A rule file written in the simple YAML format shown in the instructions
- An MCP server you already use, such as a file server, to put Toolwrit in front of
Good to know
This tool blocks and records what your AI does, but it does not undo anything, so a call that gets through still happens for real.
Install it with your AI
Add Toolwrit MCP 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 Toolwrit MCP 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 4,609 MCP servers here for you.
Who it's for
Anyone who lets an AI use tools that touch real files, money, or private data and needs clear limits and a record of what happened.
