Users run uninstall/reinstall but old state remains, then onboarding behaves unexpectedly (for example: no provider prompt, unusable install).
In practice, there are three different layers:
/usr/lib/node_modules/openclaw)/usr/bin/openclaw)~/.openclaw, for root this is /root/.openclaw)If only layer 1 is removed, old state in layer 3 can still affect onboarding.
Use this exact wipe + reinstall flow.
# 0) Stop gateway if running
openclaw gateway stop || true
# 1) Remove global package
npm uninstall -g openclaw
# 2) Remove symlink (if still there)
rm -f /usr/bin/openclaw
# 3) Remove global module dir (if still there)
rm -rf /usr/lib/node_modules/openclaw
# 4) Remove user state/config (critical for truly clean reinstall)
rm -rf /root/.openclaw
# 5) Verify removed
command -v openclaw || echo "openclaw binary removed"
ls -la /usr/lib/node_modules | grep openclaw || echo "module removed"
ls -la /root | grep .openclaw || echo "user state removed"
# 6) Reinstall
curl -fsSL https://openclaw.ai/install.sh | bash
# 7) Run onboarding explicitly
openclaw onboard
/root/.openclaw with your actual home path if not running as root.cp -a /root/.openclaw /root/.openclaw.bak.$(date +%F-%H%M%S)
Validation checklist:
command -v openclaw returns /usr/bin/openclaw after reinstallopenclaw status worksopenclaw onboard shows provider setup flow or confirms configured provider