SafeBump / update failures / plugin-target-unavailable
OpenClaw update failure: plugin-target-unavailable
plugin-target-unavailable: 69 reports, 0 still open
What it means: An enabled npm plugin has no version that resolves for the new core, or its registry metadata couldn't be read. The update refuses before it changes anything.
What to do: Retry later, pin the core with openclaw update --tag <older-version>, or disable that plugin and retry. Older updaters can refuse before the new code runs. In that case, use the manual method and then openclaw update repair.
Latest reports
- #167422 2026.9.3 · linux/x64 · 2026-10-08 · closed (completed)
- #166949 2026.9.4 · linux/x64 · 2026-10-08 · closed (completed)
- #165398 2026.9.4 · darwin/x64 · 2026-10-05 · closed (completed)
- #165352 2026.9.4 · win32/x64 · 2026-10-05 · closed (completed)
- #164599 2026.9.3 · darwin/arm64 · 2026-10-03 · closed (completed)
- #161705 2026.9.3 · win32/x64 · 2026-09-30 · closed (completed)
Stuck on 2026.9.3 or 2026.9.4? The OpenClaw maintainer's own recipe (#165706) is one manual hop to the current release, then a Doctor pass: npm install -g openclaw@2026.9.8 --allow-scripts=openclaw, then openclaw doctor --fix, then openclaw gateway restart. From 2026.9.8 onward, openclaw update handles the rest. On npm 11.15 or older, drop --allow-scripts=openclaw. Back up first with openclaw backup create --verify.
Before you retry, openclaw update status --json shows the unredacted failing step on your machine, and openclaw gateway status --deep shows what is actually serving. A retry overwrites that history.
Did an update already break your agent?
Update Rescue: $149 fixed, and you only pay once it works
Gateway stuck in health: starting, a crash loop after openclaw update, Doctor refusing your workspace, memory or CPU running away, or a Telegram, WhatsApp or Discord channel gone silent. We get you back to a stable, pinned version with your sessions, memory and channels intact, then write down the pin and a safe update routine for your setup.
- Run by an engineer who has run OpenClaw agent fleets in production for paying clients since early 2026 and has rolled back silent regressions on client machines.
- Async by default: you paste logs and run the commands we send, or share your screen. We never ask for your API keys or passwords.
- Most fixes take under 24 hours. If we can't fix it, you pay nothing.
- Pay after the fix in USDT on Arbitrum One, or by card through Whop on request (payment details).
Request a rescue (GitHub)
Request by email
Would rather keep it private? Use email. A GitHub request is public, so don't paste secrets or logs that contain tokens there.
All reason codes · Is it safe to update? Per-release verdicts