THOR memory for AI coding assistants

THOR is a memory helper for AI coding assistants. You tell it things about your project once, in your own words, and it brings those notes back at the right moment. It is handy if you keep repeating the same warnings to your assistant and it keeps forgetting them.

What this MCP server does

You tell your assistant something worth remembering, like a rule or a gotcha about your project. Your assistant passes it to THOR, which stores it on your machine. Later, when your assistant is about to touch the file or run the command that note is about, THOR puts the note in front of it right then. If the note carries a small check that still passes, THOR can refuse the matching change instead of just warning. Everything stays in one file on your computer.

Click to zoom

What you can do with it

• Save a rule or gotcha about your project once, in your own words
• Get that note back automatically when your assistant reaches the file or command it is about
• Let a note with a live check refuse a matching edit or command
• Keep each project's memory separate so rules do not leak between projects
• See a health check that names notes that have gone stale or are never read
• Let your assistant add, fix, or retire notes as you work
• Point it at a project so questions about the code get answered from the real source

What it gives back to you

It gives your assistant short notes, in plain text, at the moment they matter. Notes are capped at 300 characters, and a block of notes at four notes and 1200 characters, so what lands in the chat is small. It can also return a health check that lists notes whose checks now fail, notes pointing at nothing, and notes too crowded to ever be shown. When a note with a live check matches a change, it can refuse the change instead of just telling you about it.

Before you start

Who it's for

Developers and anyone who works with an AI coding assistant on real projects and keeps repeating the same warnings.