Instructions & context

Memory file

Also called memory files.

A memory file is a file a coding agent loads at the start of a session so instructions or notes from earlier work are present again.

Example

You write in a project CLAUDE.md that tests run with npm test. The next session loads that file, so you do not have to type the command again for Claude to see it. A chat you closed is a different record.

Why it matters

Claude Code's docs say each session begins with a fresh context window. Two kinds of file carry knowledge across sessions. CLAUDE.md files are instructions you write, for a project, for yourself, or for an organization, and Claude reads them at the start of a session. Auto memory is notes Claude writes from your corrections and preferences. Those notes live in a directory with a MEMORY.md index and one file per memory.

The docs say the first 200 lines of MEMORY.md, or the first 25KB, whichever comes first, load at the start of every conversation. To confirm a CLAUDE.md file loaded, run /context and check the list under Memory files. Claude Code treats the files as context the model reads, not as a switch that blocks an action no matter what the model decides.

How it shows up on bot.ski

On bot.ski, the Memory glossary page is the general idea of facts saved for later, and the Context and memory Learn page is how a bot uses them. This page is the file a coding agent reads. Opening this page does not write a memory file for you.

Common confusion

A memory file is not the conversation history, and it is not the whole context window. History is the saved transcript. The context window is what fits in this reply. A file on disk is only useful for the reply if the product loads it.

Sources

Updated October 3, 2026.

Cookie preferences