Claude Code memory across sessions

Claude Code remembers exactly one thing reliably across sessions: your CLAUDE.md file. It reads that file at the start of every session, so whatever you put there persists. Everything else, the conversation, the reasoning it built up about your codebase, the thing you explained two prompts ago, is session state. When the session ends, that part is gone. And it is all tied to your machine.

So Claude Code has memory, but it is the memory you wrote down on purpose, not a growing understanding of your project.

What persists, and what does not

The line is clean once you see it.

• CLAUDE.md persists. Conventions, build and test commands, architecture notes, the do-not-do list. It is a committed file, so it survives sessions and travels with the repo. This is the right home for repo-local truth.
• The session does not persist. The chat history and the context Claude Code assembled while working are not promoted into anything durable. Start a new session and you start fresh, minus whatever is in the file.
• It is per-machine. The state lives where you ran it. Move to a different machine and even the session continuity is gone.

This is a sensible design. A coding agent that silently hoarded memory across every project would be a mess. But it means the file is doing all the long-term work, and a single file has limits.

Where the file stops

CLAUDE.md is the floor, not the whole picture, and the gaps are predictable.

• It only covers one repo. Cross-team decisions, the product brief, the roadmap that lives elsewhere, none of that is in the file, so Claude Code is blind to it. Closing that gap is the point of giving Claude Code your whole project context.
• It lags reality. Someone has to remember to edit it. Until they do, it is quietly wrong.
• It is per-tool. A decision Claude Code helped you reach does not reach Cursor or ChatGPT. They keep their own state.
• It drifts against other files. The moment you also have a .cursorrules or an AGENTS.md saying roughly the same thing, they disagree, which is why keeping CLAUDE.md and .cursorrules in sync by hand is a losing game.

The fix: a shared source Claude Code reads and writes

If you want context that outlives the session and reaches every tool and teammate, keep it outside the editor in a source the tools read.

That is what a shared context layer does. BaseThread keeps a curated context graph for your team, curated, not scraped, organized into layers like Company, Products, Teams, Projects, and You, with three streams: Activity for what happened, Decisions for what you settled and why, and Tasks for what is next. Claude Code connects over MCP through the local Mac app, reads the relevant slice at the start of every session, and writes activity, decisions, and tasks back as it works.

In practice that means:

• Keep a CLAUDE.md for genuinely repo-local detail: build commands, local conventions.
• Connect Claude Code to the shared layer for everything bigger than one repo: the brief, the owners, the cross-team decisions.
• The decisions the team already logged are there, so Claude Code stops re-proposing settled choices.
• A teammate's Claude Code reads the same source, so answers agree across people, not just across your own sessions.
• When a task ships, it gets written back, so the next session, yours or theirs, starts caught up.

This is shared context for AI tools applied to Claude Code's memory. The session does not need to remember if the durable context lives where every tool can reach it.

Related reading

Does Cursor remember across sessions?

Does Cursor remember across sessions? Mostly no. Here is what Cursor keeps, what it forgets, and how shared context carries across tools and teammates.

How to give Claude Code your whole project context

Give Claude Code your whole project context, not just one repo's CLAUDE.md. Here is how a shared source keeps it current across repos, tools, and teammates.

How to keep CLAUDE.md and .cursorrules in sync

Keeping CLAUDE.md and .cursorrules in sync by hand drifts fast. Here is why, and how a shared context source keeps every tool's rules current across a team.

What is shared context for AI tools? (2026 guide)

Shared context for AI tools is the company, project, and decision background every AI reads automatically, so your whole team's tools stop guessing.