Docs / Connect / Claude Code memory

Index Claude Code memory in ModelBrain

Claude Code writes notes to its future self, and those notes are only worth something if your other assistants can read them too. The claude-code source reads a Claude Code memory folder, or a project's CLAUDE.md files, and files each note as its own memory in a partition called Claude Code. It only reads. Nothing is ever written into .claude. Available from ModelBrain 0.6.6.

Before you start

Know which folder you want. Claude Code keeps per-project memory in ~/.claude/projects/<project>/memory. You can also point the source at a project folder to read its CLAUDE.md files and rules.
Use ModelBrain 0.6.6 or later. The Add Source menu is new in that release.
Nothing is added until you add it. ModelBrain never guesses a source kind from a file extension, so a folder you already sync as a plain vault is not reindexed by this feature.

Add it

Open the ModelBrain menu, choose Vaults, then Add Source, then Claude Code memory….
Choose the memory folder, or a project folder.
ModelBrain registers the folder and indexes it in the background. A notification confirms the source was added.

From a terminal, connector_sync register <id> <path> --type claude-code does the same. For example: connector_sync register cc-notes ~/.claude/projects/my-app/memory --type claude-code.

What it reads

The source picks a layout from the folder you give it.

Memory folder. A folder named memory, or any folder that holds a MEMORY.md. Every .md file under it is read, down to 8 levels.
Project folder. CLAUDE.md, CLAUDE.local.md and .claude/CLAUDE.md at the root, plus the .md files in .claude/rules, .claude/commands and .claude/agents.
One file is one memory. Its address is the file's path inside the folder you chose.
A type value in a note's frontmatter becomes the memory's object type, so recall can filter on it. Tags in frontmatter carry over as they do for any Markdown note.
The date on each memory is the file's modified time.

What it skips

Every dot folder except .claude, and any file whose name starts with a dot or ~$.
Anything that is not .md.
A symbolic link that points outside the folder you chose.
Folders you excluded with --exclude-folder.

Limits and safety

A container file over 4 GiB is refused and recorded as a failure, not dropped. Set MODELBRAIN_CONTAINER_MAX_BYTES to change that ceiling. A single member over 32 MiB is refused the same way.

ModelBrain removes a member from memory only after it has read its container to the end. If a file cannot be parsed, every member already indexed from it stays put until the next good pass.

This source is hidden from relay callers. Anything that reaches ModelBrain through the hosted relay, which is how ChatGPT and Grok connect, cannot see what this source imported unless you add its partition (and the partition of any other source of the same kind) to that connection's allowlist yourself. Claude Code, Cursor and other local clients, which talk to ModelBrain on your own computer, can still see it.

Check that it works

Ask your assistant to run ModelBrain diagnostics. The source reports its kind, members, members_failed and last_container_bytes.
In another assistant, ask about a decision that only appears in a Claude Code memory note.

If something does not work

Nothing was indexed

Check the folder holds .md files at the level the layout expects. A project folder with no CLAUDE.md and no .claude folder has nothing to read.

Claude Code's memory path differs from the one above

Claude Code chooses its own folder names and does not document them. Browse to the folder in Finder or File Explorer and pick it directly.

Questions

How do I add Claude Code memory to ModelBrain?

Open the ModelBrain menu, choose Vaults, then Add Source, then Claude Code memory, and pick the memory folder, usually inside ~/.claude/projects. From a terminal, run connector_sync register with --type claude-code. Available from ModelBrain 0.6.6.

Does ModelBrain change my .claude folder?

No. The source only reads. It never writes into .claude or into the folder you chose.

Can ChatGPT or Grok see these notes?

No. The partition is hidden from relay callers, which is how ChatGPT and Grok connect. Local clients such as Claude Desktop, Claude Code and Cursor can read it.

Which files does it read?

In a memory folder, every .md file. In a project folder, CLAUDE.md, CLAUDE.local.md, .claude/CLAUDE.md and the .md files under .claude/rules, .claude/commands and .claude/agents.

Other sources

Each new source kind has its own page.

ChatGPT export → · Mail → · Calendar and contacts → · Bookmarks →

Connect an assistant

Sources feed the same memory these assistants read.

Claude Desktop → · Claude Code → · Cursor →