What a useful Claude Code backup covers
Anthropic documents that Claude Code stores session transcripts locally and uses CLAUDE.md plus auto memory to carry context across sessions. Those layers can sit outside Git or be split between your user profile and project folders.
A complete plan therefore separates repository backup from Claude Code continuity. It records the active Claude configuration directory, includes supported project instructions and customisations, and treats authentication material as a security boundary rather than ordinary backup content.
- The active native Windows Claude configuration directory, normally %USERPROFILE%\.claude
- Local session transcripts and auto memory under the Claude projects area
- Supported CLAUDE.md, .claude rules, commands, agents and skills
- Git repositories and original files through source control and normal file backup
A safer manual process
- 01
Confirm the owning environment
Work out whether the state belongs to native Windows or WSL. Their home directories are separate and should not be merged by assumption.
- 02
Locate the active Claude directory
Start with %USERPROFILE%\.claude and check whether CLAUDE_CONFIG_DIR changes the location.
- 03
Close Claude Code for a simple copy
A manual folder copy has no point-in-time consistency controls. Close active sessions before copying changing state.
- 04
Exclude credentials
Do not place credential files, tokens, private keys or secret-bearing configuration in an unencrypted archive or repository.
- 05
Plan storage and retention
Use a destination that can survive loss of the PC. Repeated full-folder copies store unchanged data again, so monitor capacity and remove old copies only under a deliberate retention policy.
- 06
Test away from live state
Reconstruct or inspect the copy in a separate folder before trusting it. Do not make live replacement the first recovery test.
Common mistakes
- Assuming Git contains every Claude Code transcript and user setting
- Copying credentials into a shared drive or support ticket
- Keeping the only backup on the same Windows disk
- Retaining repeated full-folder copies without monitoring how quickly storage grows
- Mixing Claude Desktop, web sessions, WSL and native Windows state
- Calling a successful copy a proven recovery without reconstructing it
What Calmbu currently claims
You can maintain this process yourself if you keep the active Claude Code locations, exclusions, storage, retention and recovery tests current. Calmbu packages that recurring work into maintained agent-aware protection, AES-256-GCM encryption, storage-efficient recovery points that reuse unchanged protected data, and required recovery checks. The public Windows installer remains gated.
See the recovery model How Calmbu verifies its claimsCommon questions
Does copying my repository back up Claude Code?
It can protect committed project instructions such as CLAUDE.md, but user-level sessions, auto memory and settings can live outside the repository. Protect both layers deliberately.
Can I back up Claude Code while it is open?
A purpose-built capture can handle changing files, but a basic manual copy cannot guarantee consistency. Close Claude Code before a manual folder copy.
Will repeated manual Claude Code backups use more storage?
They can. If every backup is another complete folder copy, unchanged local workspace data is stored again each time. Calmbu reuses unchanged protected data across recovery points, so a new point does not require another complete copy. Actual storage use still depends on how much supported workspace data changes.
Does Calmbu include Claude Code credentials?
No. Known credential files, tokens, private keys and secret-bearing values are excluded or sanitised from normal recovery points.
Official references used for Claude Code facts
Anthropic documentation establishes Claude Code behaviour. Calmbu-specific statements describe the current product contract and are not presented as Anthropic guidance.