OpenClaw v2026.11.1 Released: A Pure Housekeeping Release That Fixes the Most Annoying Bugs
On November 5, OpenClaw shipped v2026.11.1 — a rare "zero new features" release. The project defines it as a patch release: no new functionality, just fixing the three most painful problems — accepted Telegram/Discord messages being wrongly refused, failed Doctor inspections reported as successful, and Gateway configuration writes getting stuck.
Behind it: 83 pull requests, 91 direct commits, and 67 contributors.
The Three Headline Fixes
1. False message refusals. Telegram and Discord messages that had already been accepted could receive a bogus "restart refusal" while queued, or lose a fallback reply; queued Telegram replies could fail after a previous conversation turn ended. In a product where instant messaging is the main artery, all three (PRs #179200, #179385, #189457) were keenly felt.
2. Doctor stops lying. Previously, a blocked inspection or incomplete repair still exited with code 0 — automation had no way to notice. The new version reports failure with a nonzero exit code (scripts parsing doctor --json should adapt), and unattended Doctor runs no longer install missing plugins silently; you must pass --fix, --repair or --yes explicitly.
3. Configuration writes unstuck. Gateway settings changes could hang forever waiting to apply. Fixed, and channel restarts are more reliable as well (adaptation of #192701).
The Update-Recovery System Keeps Hardening
OpenClaw's recent releases share a theme — "make upgrading less of a gamble" — and this one continues:
- SQLite lock retries: Gateway startup retries temporary SQLite locks instead of stopping immediately
- Migration tolerance: Doctor skips legacy data with malformed UTF-8 (preserving original bytes for recovery), keeps invalid heartbeat settings and tells you what to fix, and can reunite stores split by older relative paths
- Node pre-checks: Git updates check the target Node requirement before staging; services started from root can find a compatible runtime
- Windows fixes: update recovery keeps the scheduled task available until the Gateway is ready; an update can restart a service it stopped after a recoverable Doctor failure
- Honest reporting: intentional skips in automatic updates now report as
skippedinstead offailure - Users on 2026.10.7–10.9 can invoke a fresh Doctor during failed-update recovery
Polishing Every Module
Messaging: iMessage gains PostgreSQL-backed state; Slack relay connections and stalled Tlon streams can reconnect; Feishu keeps streamed text across tool calls and formats action cards correctly; Google Chat sends remote attachment links with captions; queue-overflow summaries keep sender and conversation context.
Models and providers: Gemini Flash text models on Google AI Studio default to the Interactions API (more consistent key rotation and sampling); Anthropic API-key setup accepts a workspace ID; stable tool definitions improve prompt-cache reuse across turns; rate-limited profiles honor the provider's wait time; Codex conversations retain thread history across turns and can recover an answer after a transient retry without re-running completed tools.
Desktop and mobile: the Mac app no longer overwrites real credentials with redaction placeholders when a save fails — a genuine data-loss bug; Voice Wake avoids duplicate sends; Android Talk lets accepted work finish and its answer appear in chat; long spoken replies queue for playback.
Security: Discord attachment downloads no longer inherit broad browser private-network permissions; ClawHub restricts private-network downloads and redirects (SSRF hardening); streamed replies no longer leak split or nested internal reasoning blocks; Fish Audio and Gradium credentials are redacted from provider errors.
Why a "Zero-Feature" Release Matters
v2026.11.1 is a good sample for gauging the project's maturity: a project with 80,000+ forks willing to devote an entire release to stability work, whose fix list is full of items like "report failure honestly," "preserve original bytes," and "be recoverable" — details with zero demo value that nonetheless determine production trust.
Source: OpenClaw official release notes (docs.openclaw.ai/releases/2026.11.1), GitHub Release
