Upgrade watch
OpenClaw Upgrade Watch
A standing operator page for OpenClaw upgrade recommendations: what we are running, what changed, what risks showed up, and whether to hold, watch, or test.
Keep production-like baselines on 2026.4.23. Treat 2026.6.9 stable and 2026.6.10-beta.1 as watch/lab-only candidates until canary smoke tests pass.
Gateway loopback-only, token auth, stable Telegram path, status via JSON wrapper preferred.
Latest stable found: 2026.6.9. Latest prerelease found: 2026.6.10-beta.1. Local preflight is clean, but source risk keeps production on hold.
Why we are cautious
- Production assistants are infrastructure. A broken gateway or channel connector means the operator loses the cockpit.
- Channel regressions matter. Telegram and messaging paths should be tested before any upgrade is considered safe.
- Config mistakes should be recoverable. Keep backups, restore packs, and rejected-config rollback notes.
- Updates need a pipeline. Scout release notes and community signal, classify
HOLD/WATCH/SAFE TO TEST, then test messaging, gateway, sessions, status checks, and critical workflows before declaring a release good. - Local runtimes can drift. Avoid relying on auto-updating local model services for boot/heartbeat paths unless intentionally approved.
- Skills are supply chain. Treat public skill marketplaces as useful but risky; prefer auditable code and narrow permissions.
Upgrade decision model
HOLD
Use when there are channel regressions, unclear release notes, auth/provider changes, serious Windows issues, or insufficient backup/test coverage.
WATCH
Use when a release looks interesting but needs community signal, Discord/GitHub review, or a low-risk test machine before touching production.
SAFE TO TEST
Use only after backup, changelog review, known connector checks, and a clear rollback path. Test first; production later.
Latest scan notes - 2026-06-21
Decision: HOLD production / WATCH for lab canary. The known-good production baseline remains 2026.4.23. The latest npm stable release found was 2026.6.9; the latest GitHub prerelease found was 2026.6.10-beta.1.
Why: The June release train contains useful fixes for stateful operator paths: pending subagent completion announcements, non-empty chat histories, follow-up drains, Codex approval/session flows, Telegram progress/thread delivery, status workflows, loopback-scoped network checks, and plugin/skill behavior. Those are exactly the surfaces that matter to a phone-first operator cockpit.
Risk signal: Same-day prerelease signal is still too fresh for production. A Windows beta-upgrade report documented plugin version drift where some plugins stayed pinned to 2026.6.9 after upgrading to 2026.6.10-beta.1, with repair-command guidance still being refined. Channel and account-migration changes also need soak time.
Local posture: A redacted readiness preflight for 2026.6.9 passed the local checks. That makes a backed-up canary reasonable, not a production upgrade.
Reopen trigger: plan a non-production canary for 2026.6.9 first, then consider beta only if its fixes are specifically needed. Minimum smoke tests: gateway status, Telegram inbound/outbound, cross-session replies, /stop, subagent completion announcements, plugin drift/doctor output, heartbeat/no-login startup, and rollback.