Molt

Snapshot and back up OpenClaw brain files (AGENTS.md, SOUL.md, MEMORY.md, memory/, etc.) to an offsite git repository — like a lobster shedding its shell, leaving behind a perfect copy before growing. Use when asked to back up the brain, run a molt, snapshot memory offsite, or schedule/manage brain backups. Also use when setting up or configuring the molt cron job.

Snapshots OpenClaw brain files to a remote git repo — a perfect copy of who you were before you grow. Idempotent — only commits when something changed.

Security Model

molt sends files only to a git repository you control — no third-party servers, no telemetry, no hidden destinations. The target repo is set explicitly by you via MOLT_REPO_URL or --repo-url.

Sensitive fields are redacted before backup using openclaw config get, which is the authoritative redaction source built into OpenClaw itself. The redacted config is the only file containing config data — raw openclaw.json is never copied.

This is an intentional, user-initiated backup tool. You own the destination.

Requirements

• git — must be installed and on $PATH
• Git authentication — the script uses git clone / git push over HTTPS or SSH. You need credentials configured for the remote:
  • SSH (recommended): add your SSH key to GitHub/GitLab; use a [e-mail]:... repo URL
  • HTTPS with token: configure a credential helper or use a token in the URL (https://<token>@github.com/...)
  • gh CLI is not required — molt uses plain git only
• rsync — used for memory/ sync; standard on macOS and most Linux distros
• Write access to the remote repo — the pushing account needs push rights; a dedicated private repo is recommended

No other dependencies. No OpenClaw-specific CLI tools needed at runtime.

Zero-to-Working Setup

1. Create a private remote repo (GitHub, GitLab, etc.) — e.g. your-org/your-brain
2. Ensure git auth works for that repo from the machine running OpenClaw:

   bash

   ssh -T [e-mail]   # for SSH — should greet you by name
   

3. Run molt once with the repo URL:

   bash

   MOLT_REPO_URL=[e-mail]:your-org/your-brain.git \
     ~/.openclaw/workspace/marv-skills/molt/scripts/molt.sh
   

   This clones the repo to ~/.openclaw/molt, copies brain files, commits, and pushes.
4. Schedule it — add a cron job or ask your assistant to set one up via openclaw cron.

After the first run, the MOLT_REPO_URL env var is no longer needed (the local clone already knows its remote).

What Gets Backed Up

• AGENTS.md, SOUL.md, TOOLS.md, IDENTITY.md, USER.md, HEARTBEAT.md, MEMORY.md
• memory/ directory (daily notes)
• cron-jobs/ — exported live from openclaw cron list at backup time (JSON + Markdown summary)
• config-redacted.json — OpenClaw config with all secrets/tokens replaced by [REDACTED]

Config

Set via environment variables or pass as flags:

Env var	Flag	Default	Description
MOLT_WORKSPACE	--workspace	~/.openclaw/workspace	Source workspace
MOLT_DIR	--backup-dir	~/.openclaw/molt	Local backup repo clone
MOLT_REPO_URL	--repo-url	(required on first run)	Remote git repo URL
MOLT_EXTRA_DIRS	--extra-dirs	(none)	Comma-separated workspace subdirs to include (e.g. scripts,notes)

Running

bash

# First run (clones the repo):
MOLT_REPO_URL=https://github.com/your-org/your-brain-repo.git \
  scripts/molt.sh

# Subsequent runs (repo already cloned):
scripts/molt.sh

# Dry run (see what would change without pushing):
scripts/molt.sh --dry-run

Scheduling

To schedule automatic backups, create a cron job pointing at this script. Example for every 6 hours:

text

0 */6 * * * MOLT_DIR=~/.openclaw/molt ~/.openclaw/workspace/marv-skills/molt/scripts/molt.sh >> /tmp/molt.log 2>&1

Or use openclaw cron to schedule via the assistant.

First-Time Setup

1. Create the remote repo (GitHub, GitLab, etc.)
2. Set MOLT_REPO_URL to the repo URL
3. Run the script once — it will clone and push
4. On subsequent runs, only env var MOLT_DIR is needed (defaults to ~/.openclaw/molt)