Adapted community discussion: Paraphrased and condensed from an anonymized public OpenClaw Discord support conversation from September 2026. Technical guidance was reviewed for OpenClaw 2026.9.3.
I updated my Mac and OpenClaw no longer works. The CLI reports missing multi-agent ownership, the Gateway refuses local connections, and Doctor has several pending state migrations.
Confirm the version and agent IDs first. An unscoped status command can fail in a multi-agent setup even when the CLI itself works. Use openclaw agents list --bindings, target diagnostic commands with the intended agent ID, and let Doctor report the migration blockers before using any force option.
Doctor refuses to mutate shared SQLite state because a managed Gateway process still owns the directory. The service looks loaded, but it never opens the port.
Stop the managed Gateway cleanly before state repair. When foreground startup identifies a legacy session store, use the dedicated migration in dry-run mode, then import only the agent named by the blocker:
openclaw doctor --session-sqlite dry-run --session-sqlite-agent <agent-id>
openclaw doctor --session-sqlite import --session-sqlite-agent <agent-id>
Do not delete or move sessions.json. Repeat only for another agent if foreground startup names it explicitly.
The targeted session imports complete, and the Gateway can run in the foreground. The remaining problem is that every Telegram account stays disconnected because the legacy exec-approvals migration still blocks channel startup.
Preserve the approvals file because it contains the policy behind host allow prompts. If the installed stable build cannot converge that migration and a current supported update contains the fix, preview and apply that update, then stop the Gateway, run Doctor, and start it again. Ensure several gigabytes of disk space are free before retrying an update that previously failed for capacity. Verify with openclaw channels status --probe.
After freeing disk space, the update and Gateway restart completed. The Gateway is reachable and all six Telegram accounts report running, connected, and working.
That is a clean recovery. The newer build completed the migration, the Gateway is healthy, and each configured Telegram transport is connected. Keep the preserved approvals and pre-update backup until the recovery has been stable for a while.