CLAUDE.md

Last update: September 30, 2026

In our last lesson, we learned about the importance of context and the memory-capturing ability of LLMs. If we provide context in our chats, we can expect better outputs. If we save a memory, it will persist across our chats.

Later in this series, you’re going to learn how you can drop a file into a folder, type three words, and have Claude go to work on that file. The following will happen: the file will get analyzed, logged, moved, and then a new file will be created. And it all happens because we’ll have set up a CLAUDE.md instructions file.

CLAUDE.md

First: you’ve seen CLAUDE before. But what is .md? .md stands for Markdown, a plain-text computer language that LLMs love. A .md file is just a text file. Same thing as a .doc[x] or .txt file, just more lightweight.

CLAUDE.md is an instructions file. It generally lives inside of a folder that you connect Claude Code to. (We’re going to see this in action later, when we’re using VS Code as our main Claude interface.) It helps Claude understand what to do and how to do it.

This is going to be a lot easier to follow when we jump into it live. Hang on for me in the meantime. Read on:

What’s important to know is that CLAUDE.md is the first file that Claude Code will read when looking at a folder. So we want to set it up as a sort of “guide” for Claude to reference. If we have child folders, for instance, we should tell the Claude what their purposes are and when to interact with them. If we ask Claude to do something… this “guide” tells Claude what and how something should be done.

If you have multiple folders, you can have multiple CLAUDE.md files (one in each folder). Each time Claude steps into a folder, it will read that CLAUDE.md file.

We’ll demonstrate this live in the next post.

The key takeaways

• Claude has access to our files through VS Code
• CLAUDE.md is an instructions file that Claude reads when it opens a folder
• We can customize CLAUDE.md to give it context, memories, instructions, etc