Claude Code memory vs team memory: when CLAUDE.md and auto memory stop being enough
By Soumen Sarkar · October 2026
Comparison. If you work alone, or each person owns their own repository, Claude Code's CLAUDE.md, auto memory and Projects are enough: keep using them. If a second person's sessions change the same product, a call made in your session has to reach theirs the same hour, and that's what shared team memory like Campfyrr adds. Most teams start with the first and grow into the second.
What are you trying to fix?
If you work alone, or each person owns their own repository, Claude Code's built-in memory is enough. If two people's sessions change the same product, you need a call made in one person's session to reach the other's while they work. That's shared team memory, and Campfyrr is one. Most teams start with the first and grow into the second.
- "I keep telling Claude the same thing." Built in. Auto memory is notes Claude writes for itself from your corrections and preferences.
- "Every session should follow our standing rules." Built in. The project CLAUDE.md is "shared with your team through version control" (memory docs).
- "My parallel threads should share what they learned." Built in. Projects give one person's threads a shared memory of "requirements, decisions, and pitfalls", and cross-session messaging warns your other sessions when one breaks what another builds on.
- "My teammate's session just undid a call my session made this morning." That's the one the built-ins don't reach. Shared team memory fixes it.
The same day, with each
Here's a scene, not a customer. Ana and Ben are developers on one shop, with a PM.
With the built-ins. For a month Ana works alone, and it's enough: CLAUDE.md holds the rules, auto memory her habits, and her project's threads share decisions. Then Ben joins. At ten, Ana's session settles with the PM that refunds go to store credit, not back to the card. At eleven, Ben's session starts on the refund emails. Nothing carries the call across. Auto memory "is machine-local", so it stays on Ana's laptop. "A project belongs to one user. You can't share a project or its threads with another user" (Projects docs). Cross-session messaging reaches "your other Claude Code sessions", and Ben's aren't Ana's. CLAUDE.md does cross people, but only after someone edits and merges it. Ben's session writes "your refund is on its way to your card." The reviewer catches it, or a customer does.
With Campfyrr on top. Ana's call at ten is recorded as a decision with the PM's name and the reason. At eleven, the first time Ben's session opens the refund code, the hook adds what the team knows there: the store-credit decision, and Ana's work in progress on the refund flow, so he builds on it rather than around it. The emails say store credit. Nobody ran a command or relayed anything by hand.
What's different by evening: with the built-ins, each person's sessions were consistent with themselves. With Campfyrr, they were consistent with each other. See what a session gets.
When Claude Code's built-in memory is the better choice
- You work alone, or each person owns a separate repository. Nothing needs to cross between people, and the built-ins do one person's job well.
- Your calls are standing rules that change rarely. Put them in CLAUDE.md, where they're reviewed like code.
- You want one person's parallel work run for you. Projects run parallel cloud threads, and agent teams let one lead session start teammates with a shared task list. Campfyrr runs no threads and spawns no agents. Note that Projects are in beta on Pro and Max and "aren't available on Team or Enterprise plans yet", and agent teams are experimental and off by default.
- You'd rather wait for Anthropic. An open request, Shared Team Memory for Claude Code, says "Claude Code's memory system is individual-only." If you can wait, watch that issue.
What it costs
The built-ins cost nothing beyond the Claude plan you already pay for. Campfyrr Team is $9 per developer a month, minimum three ($7 billed yearly), and business teammates who read and answer are free. For 5 developers and 3 business teammates, that's 5 times $9, which is $45 a month on top of your Claude plans, or $35 billed yearly. Everything is free during early access (pricing).
The cost that matters more is who does what each day. With the built-ins alone, someone has to notice each call that matters to others, edit CLAUDE.md and get it merged, or tell teammates by hand. That relaying is the work that grows with every person you add (sessions don't know about each other). With Campfyrr nobody runs a command; people answer the assumptions sent to them, from a link. You still keep CLAUDE.md for stable rules. Campfyrr never stores your code (trust), and it doesn't run tests or replace CI. If you're already wiring Claude to every system by hand, read when the glue becomes the project first.
Verdict
Choose the built-ins if you work alone or each person owns their own repository. CLAUDE.md, auto memory and Projects are enough.
Choose Campfyrr if two or more people's sessions change the same product, and calls made in one session need to reach the others the same day.
Use both if you're a team, which is most readers. Keep stable rules in CLAUDE.md and habits in auto memory, and let Campfyrr carry the live decisions between people. Our own repository's CLAUDE.md says "The session brief arrives on its own (SessionStart hook). Read it before starting." Setup is in shared team memory for Claude Code; for agents started on GitHub, see Campfyrr vs GitHub Copilot Memory, and for the whole field, our research.
The problem this answers: Sessions don't know about each other: "My teammates' Claude Code sessions don't know what each other are doing."
The principle behind this: Answer it yourself: Look first. If the answer isn't there, assume it or decide it.
Move as one at agent speed. Nobody waits. Add Campfyrr to Claude: it briefs every agent before it touches the code, keeps assumptions and handoffs where the team can see them, and never takes your code (what it keeps).