OpenClaw patches certificate risk and startup failures in security release

OpenClaw has released v2026.10.6 to address the eMudhra certificate compromise and failures that could prevent installation, updates or startup. The patch adds no new features. Its most immediate instruction goes beyond updating OpenClaw: users who manage Node themselves must upgrade that software before starting OpenClaw, according to the release notes.

The Node requirement comes first

OpenClaw specifies separate minimum versions for the two Node branches named in its security instructions. For users managing Node themselves, the required versions are:

• Node 24: upgrade to 24.22.1 or newer within that branch.
• Node 26: upgrade to 26.12.1 or newer within that branch.

Installation fixes cover several different ways setup could fail. On Windows, false offline-maintenance messages and database errors could block a fresh setup, the Doctor repair tool or the Gateway from starting. A separate fix allows Windows installations using Bun to finish when an optional cleanup step fails.

Linux service setup also gets a correction for commands that dropped empty arguments. Other setup changes narrow dependency checks to the installers actually needed and fix Codex imports that waited too long—or gave up too early—when the computer’s clock changed.

Updates gain clearer stopping rules

The release’s update changes distinguish between failures that can carry a warning and conditions that must halt work. Configuration repair will no longer change files if it cannot save the originals first. Users must resolve the reported lock or storage problem before trying again.

Updates also stop when they cannot confirm which Gateway is running. OpenClaw instructs users to stop it through Task Scheduler, the terminal that launched it or the Gateway stop command, then retry. Shutdown now waits for pending cleanup, with extra time for Windows to confirm that the process has stopped.

Some fixes cannot rescue an older updater automatically. Windows source updates could get trapped copying through dependency links; affected users may need one manual source update to obtain the correction. Older updaters already stuck retrying a package installation that the database refused may also require manual recovery.

After a manual upgrade, the status command no longer recommends an outdated update that could undo the selected Gateway version. That change applies when status confirms a different, healthy Gateway and the advice did not specify an explicit update target. Other update and rollback warnings remain.

Recovery does not mean resuming everything

The patch fixes dead workers blocking a session’s next turn after a confirmed reboot. It does not resume the interrupted turn. The Gateway must compare saved and current boot identities before releasing those workers’ slots; if that information is older or unreadable, it continues waiting. Container cleanup remains a separate requirement.

Database recovery has similarly specific boundaries. Doctor can repair a database that a native app created first only when it is confirmed empty. A different repair addresses older Mac profiles missing a deletion-history table, but requires a verified full backup of the database and its WAL file. Items awaiting deletion may return if their deletion records were lost.

For users whose Gateway stopped starting after upgrading from 10.4 to 10.5 because of an older built-in agent database, Doctor now offers a repair. OpenClaw says to run it only after completing the backups and checks in its database integrity and recovery guidance. Healthy agents starting does not prove that the affected database was repaired.