How shared memory between AI assistants works
See what a shared memory connection can provide, what it cannot transfer, and how to test access with a harmless project detail.
Understand AI memory 4 min read
Yes, but it keeps different context in different places. The Claude apps save memory from your chats, and each Project gets its own. Projects also hold files and instructions you add, and Claude Code reads CLAUDE.md files from your computer when a session starts.
Claude saves memory as topics while you chat, such as your role, your projects and how you like answers written. You can also say “remember this.” Each Project has its own memory, kept apart from other Projects and from chats outside them. Memory is on by default on Free, Pro and Max plans. On Team and Enterprise plans, an owner has to turn it on.
Settings > Memory lists every topic to edit or delete, or you can ask Claude in a chat to forget something. Pausing keeps memory but stops Claude from using or adding to it. Resetting deletes all of it, Project memories included, for good. Incognito chats are not saved to memory.
On paid plans, Claude can also search your past chats when you ask about one. The search shows up as a tool call that links to the chats it used. Inside a Project, it covers only that Project’s chats.
Memory is what Claude picks up from your chats. A Project holds what you choose to give it. Project knowledge is for reference material: documents, text or code that Claude uses in every chat in that Project. Project instructions set how Claude behaves there and apply only inside the Project. Claude cannot see a Project’s name or description, so put what it needs in the files or instructions.
Each Claude Code session starts with a fresh context window. Two kinds of file carry context between sessions, and both load when a session starts. CLAUDE.md files hold instructions you write. Auto memory is notes Claude writes for itself, kept by default under ~/.claude/projects/ on that machine only.
A CLAUDE.md at the root of a repo, or in its .claude folder, is shared with your team through version control. Your own preferences for every project go in ~/.claude/CLAUDE.md, and personal notes for one project go in CLAUDE.local.md. Files in the folder where you start Claude and the folders above it load at launch. One in a subfolder loads when Claude reads files there.
Run /memory to open these files or turn auto memory off, and /context to see what loaded. Developers on the Claude API have a separate memory tool, which lets Claude read and write memory files that their own app stores.
TAKE THIS WITH YOU
An illustrative template. Replace the details with material you are allowed to share.
A rule for every chat (“use British spelling”): Instructions for Claude, in Settings Facts about you and your work: memory; review them in Settings > Memory Reference files for one piece of work: that Project’s knowledge How Claude should act inside one Project: project instructions A chat you do not want remembered: an incognito chat Build commands and rules for a repo: CLAUDE.md, committed to the repo Your own coding habits in every repo: ~/.claude/CLAUDE.md Context ChatGPT or Grok also needs: nothing in Claude keeps it in sync; paste a checked brief or use a shared memory tool
USING THIS WITH DRAY
Everything above stays inside Claude. Dray is for memory shared across assistants: its Chrome extension captures supported web chats, and assistants you connect through MCP can read the result. It is separate from Claude’s memory, and an MCP connection to a CLI agent such as Claude Code saves nothing by itself. Running dray setup saves its sessions too.
Available sources and connections depend on your setup. Read the privacy policy before granting access.
Product details are grounded in the guides below. Example projects and prompts are illustrative, not customer results.