cerveau boot <name> — launch Claude Code inside a brain from anywhere. Passes extra args through (e.g. cerveau boot MyApp --resume).

cerveau dir brain|code <name> — print the absolute path to a brain or its codebase directory. Pipe-friendly, single-line output.

cerveau cd brain|code <name> — navigate to brain or codebase directory via shell wrapper.

cerveau completion zsh|bash — shell completion scripts with tab-tab support for commands, brain names, packages, tags, and orgs. Includes the shell wrapper for cerveau cd.

cerveau marketplace list filters — filter packages by free text, --tag, or --org (case-insensitive). Example: cerveau marketplace list --org _local_.

cerveau backup [flags] — create a backup archive of ~/.cerveau/, ~/.claude/, and/or MDPlanner data. Flags: --all (default), --cerveau, --mdplanner, --claude, -o <path>.

cerveau restore <archive> [flags] — restore from a backup archive with confirmation prompt. Selective restore via --cerveau, --mdplanner, --claude.

cerveau spawn output now recommends cerveau boot instead of cd && claude.

install.sh reworked: extracts to /tmp, copies only runtime files to ~/.cerveau/, cleans up staging on exit.

cerveau update reworked: allowlist-based — only runtime files are updated, user data (_brains_/, data/, _packages_/_local_/, .env, brains.json) is never touched.

Non-runtime files (cmd/, docs/, go.mod, install.sh, LICENSE, README.md) are no longer copied to ~/.cerveau/.

Documentation updated: quick-start, first-brain, CLI reference, marketplace, and landing page all use cerveau boot as the primary launch method.