Start ClawedBack

Run this whenever starting or restarting Claude Code in a ClawedBack installation. It ensures everything is up and running.

Step 1: Locate the project and config

PROJECT_ROOT="$(pwd)"
cat "$PROJECT_ROOT/data/clawedback.json" 2>/dev/null || echo "NO_CONFIG"

If no config exists, tell the user to run /oc-setup first (first-time install) or /oc-resume (if config was lost and needs rebuilding). STOP.

If config exists, read the mode, python path, server command, and port from it.

Step 2: Ensure the web server is running

curl -sk http://localhost:${OC_PORT:-8080}/api/health 2>/dev/null

If health check succeeds — server is already running, skip to Step 3.

If server is down — start it using the saved server_cmd from the config:

pkill -f "python3 main.py" 2>/dev/null || true
sleep 1

Then start the server (read the exact command from clawedback.json's server_cmd field):

nohup $SERVER_CMD > $PROJECT_ROOT/data/logs/server.log 2>&1 &
sleep 2 && pgrep -f "python3 main.py" | tail -1 > $PROJECT_ROOT/data/server.pid

Verify:

curl -sk http://localhost:${OC_PORT:-8080}/api/health

If still failing, show the log:

tail -20 $PROJECT_ROOT/data/logs/server.log

Step 3: Create the polling cron job

CronCreate with cron "*/1 * * * *" and prompt "/oc-poll"

This schedules the heartbeat. No need to check if one already exists — CronCreate handles that.

Step 4: Run one poll cycle

Run /oc-poll once immediately to process any messages that came in while Claude was offline.

Step 5: Confirm

Tell the user:

ClawedBack started.
Server: [URL from config]
Polling: active (every 1 minute)

Keep it short. They know what it does.