Features

Feature	Description
Multi-Brain	One protocol, many projects. Each brain gets selective rules, hooks, and agents — no duplication, no drift.
Selective Loading	Declare exactly which stacks, practices, and workflows each brain needs. Only those rules load into context.
MDPlanner Integration	Single source of truth for tasks, notes, decisions, and progress. Claude reads and writes it every session.
Hooks Enforcement	Boot reminders, commit validation, checkpoint checks, and progress gates. The protocol runs automatically.
Protocol-Driven Workflow	Four phases. Boot → Work → Write Back → Close. Zero drift between sessions.
Zero Footprint	No files added to your project repos. The brain lives outside your code. Your codebase stays clean.
Bring Your Own Rules	No rules ship by default. Generate stack, practice, and workflow rules with a single Claude prompt.
Marketplace	Browse and install community packages (workflows, practices, agents) into any brain with one command. Filter by text, tag, or org.
Shell Completions	Tab-tab support for all commands, brain names, packages, tags, and orgs. Includes `cerveau cd` shell wrapper.
Boot from Anywhere	`cerveau boot <name>` launches Claude Code in any brain from any directory. No `cd` needed.
Backup & Restore	`cerveau backup` archives `~/.cerveau/`, `~/.claude/`, and MDPlanner data. Selective with `--cerveau`, `--mdplanner`, `--claude`.
Auto-Update	Version check on session start. Update the protocol in place — `.env`, brains, and config are preserved.
One-Line Install	`curl -fsSL https://cerveau.dev/install.sh \| bash` — installs protocol, starts MDPlanner, wires MCP.
Open Source	AGPL-3.0. Self-host everything. MDPlanner runs in a container. No external dependencies required.
Agent Support	Define custom agents in YAML. Declare which agents each brain loads. Agents live in the protocol.
Skills Support	Reusable skill definitions (slash commands) shared across all brains from the protocol.