OpenClaw 2026.4.22 dropped April 22 with a lot of headline features: full xAI integration (Grok image generation, 6 TTS voices, real-time STT), new local TUI mode, /models add from chat, WhatsApp native reply quoting. Those are real additions worth knowing about.
But if you're running OpenClaw as a production agent — not just experimenting with it — the more important story is the stability fixes. A 24/7 agent that duplicates messages, loses config on restart, or crashes when a plugin throws an unexpected error is worse than no agent at all. This release addressed several of those chronic failure modes.
Here's what actually changed for reliability, and what it means in practice.
The Three Core Problem Areas This Release Fixed
The 2026.4.22 changelog addresses three categories of production pain:
- Message delivery reliability — duplicate sends, dropped messages, wrong routing
- Session and config stability — settings lost on restart, config merge conflicts causing unexpected behavior
- Plugin and service compatibility — crashes and silent failures when third-party plugins or integrations behaved unexpectedly
Message Delivery Fixes
WhatsApp message duplication
A persistent issue for WhatsApp users: certain message types were delivered multiple times, with no clear trigger. 2026.4.22 fixes the underlying deduplication logic. If you've been working around this by adding delays or avoiding certain response formats on WhatsApp, you can remove those workarounds.
WhatsApp native reply quoting
OpenClaw can now reference the specific message it's replying to in WhatsApp conversations — the native "quote reply" behavior that makes threaded WhatsApp conversations readable. Previously, replies appeared as standalone messages with no visual connection to the original. This matters for group chat deployments where multiple people are talking to the same agent.
Slack thread alias preservation
Slack thread continuity was breaking in certain configurations — replies going to the wrong thread or losing thread context after restart. Fixed in this release. If you're using OpenClaw for Slack-based team operations, restart after updating to reset the thread state correctly.
Session and Config Stability Fixes
Config merge logic overhaul
One of the most frustrating long-running issues: settings changed in openclaw.json sometimes failing to apply correctly, or reverting after gateway restart. The root cause was a config merge conflict when multiple sources (base config, plugin config, workspace config) provided overlapping keys. 2026.4.22 rewrites the merge logic to be deterministic and order-of-precedence based.
Practical implication: if you've been applying config changes and seeing them not stick, update to 4.22 and re-apply. The new merge logic should hold.
Onboarding TUI startup speed
The openclaw configure wizard was significantly slower than it needed to be — it was running full gateway probe logic even when the gateway wasn't available. Fixed: configure now skips the gateway probe on startup, so the onboarding TUI reaches its first prompt immediately.
Model catalog on fresh install
Fresh installs now show bundled provider catalog rows (including Kimi K2.6 rows for Moonshot, OpenRouter, and Vercel AI Gateway) in models list --all before any auth is configured. Previously, the model list was empty until you'd set up API keys, which made initial model selection confusing.
Plugin and Service Compatibility Fixes
Plugin discovery hang (lossless-claw)
A strict-match contract added in 2026.4.14 broke plugins whose internal engine ID didn't exactly equal their registered slot ID — including lossless-claw and several other community plugins. This produced repeated info.id must match registered id failures on every turn without crashing, just silently degrading performance. Fixed in 4.22.
Browser ref validation speed
Browser automation tool calls were taking longer than expected due to ref validation overhead. The validation logic has been optimized — no behavior change, just faster execution for any workflow that uses the browser tool heavily.
Owner command security enforcement
Commands restricted to the owner role were not consistently enforcing that restriction in certain edge cases — specifically when channel routing passed through an intermediary plugin. This is a security fix, not just a stability fix. Patched in 4.22.
The New Features Worth Knowing
For completeness, the headline additions:
Full xAI / Grok integration
Image generation (grok-imagine), 6 TTS voices (MP3/WAV/PCM/G.711), real-time and batch STT. Configure with xai/grok-3 as model.
Local TUI embedded mode
Run OpenClaw in a terminal UI without starting the gateway. Good for local-only testing and development workflows.
/models add from chat
Add new models live during a session without restarting: /models add moonshot/kimi-k2.5. Useful when switching providers mid-session.
Per-group WhatsApp system prompts
Different SOUL.md behavior per WhatsApp group. Useful for teams using one OpenClaw instance across multiple group contexts.
How to Update
Verify the version with openclaw --version — you should see 2026.4.22 or newer. After restarting, check your config loaded correctly with openclaw status.
If you're on any version before 2026.3.31: Update immediately regardless of these features. CVE-2026-41352 (CVSS 8.8 RCE) and CVE-2026-41349 (CVSS 8.8 consent bypass) are both fixed in the 2026.3.x patch series. Running unpatched is a meaningful risk if your gateway is reachable from outside your local network.
Who Should Prioritize This Update
This update is worth prioritizing if you're:
- Running WhatsApp integrations (duplication fix + native replies)
- Using Slack for team operations (thread continuity fix)
- Using lossless-claw or any plugin that was throwing
info.id must matcherrors - Experiencing config changes not applying correctly after restart
- Interested in xAI Grok for image generation or TTS
- On any version before 2026.3.31 (security — do this immediately)
If none of those apply, it's still a good routine update — the config merge and browser validation improvements benefit every deployment even if you haven't noticed the symptoms.
Running OpenClaw and Not Sure If It's Configured Correctly?
ClawReady's $49 audit reviews your full setup: version, config, gateway binding, security posture, and memory architecture. You get a written report and a prioritized fix list. Takes 24 hours.
Book an Audit →