Team memory for Claude Code

What Claude Code remembers today

Every Claude Code session starts with a fresh context window. Two built-in mechanisms carry knowledge into it:

• CLAUDE.md files: instructions you write, at organization, user, project (./CLAUDE.md, shared through git) or local scope. Claude Code's docs suggest keeping each one under 200 lines.
• Auto memory: notes Claude writes itself from your corrections and preferences, kept per repository. The first 200 lines or 25KB load into each session.

Both load at the start of every session, as context rather than enforced configuration.

Where built-in memory stops

• It is loaded, not searched. Everything in CLAUDE.md enters the context window at launch, and auto memory is cut at 200 lines, so both have to stay small.
• Auto memory is yours alone. What Claude learns on your machine does not reach a teammate's session.
• CLAUDE.md is written by hand. It holds the instructions someone wrote down, not the decision your agent reached yesterday or the root cause it found.
• It stays in Claude Code. A fix found in Claude Code does not reach the same repository in Cursor or Codex.

CLAUDE.md is still the right place for build commands and standing rules. Most teams use both.

What Recallium adds to Claude Code

• Typed memories as the work lands: designs, decisions, root causes, fixes and runbooks, each linked to its files
• Search instead of load: before an edit, Claude searches by topic and file path and expands only the results it needs
• Shared with your team: on Cloud Pro, every teammate's agent reads and writes the repository's memory
• Shared with your other agents: the same memory in Codex, Cursor, GitHub Copilot and any MCP client
• Handoffs: workstreams, working state and session recap carry an effort across sessions and teammates
• Commit trailers: a Recallium-Memory: trailer links each commit to the memories that explain it

Add Recallium to Claude Code

1. Run the installer: npx -y recallium install finds Claude Code and your other agents, then adds the MCP server, the Recallium guidance and the supported hooks
2. Authenticate: in Claude Code, run /mcp, select recallium and choose Authenticate
3. Open a repository and work: Recallium names the project after the repository and loads its context at the start of each session

Full steps are in the quickstart. Recallium Cloud is in a closed pilot; join the waitlist, or self-host the community edition.

Questions

Does Claude Code have memory?

Yes, two kinds: CLAUDE.md files you write, and auto memory Claude writes itself, per repository on your machine. Both load at the start of each session. Neither is shared automatically with teammates or other agents; an MCP memory server such as Recallium adds that.

CLAUDE.md, auto memory or Recallium: which should hold what?

CLAUDE.md for standing instructions and build commands. Auto memory for your own preferences. Recallium for the team's record: why the code is the way it is, what was tried, and what fixed it.

Can my team share Claude Code memory?

A project CLAUDE.md is shared through git. Auto memory is not. Recallium Cloud Pro shares the memory itself: what each teammate's agent learns in a repository is available to everyone on it.