OpenClaw v2026.5.5: 60+ Fixes for WhatsApp, Feishu, Discord & Gateway Reliability
OpenClaw v2026.5.5 is a comprehensive patch release that resolves over 60 issues spanning every major sales channel, the Control UI, gateway reliability, provider integrations, and the plugin system. If your B2B SDR agents run in production across WhatsApp, Discord, Feishu, LINE, or Slack, this update removes a significant class of silent failures that were costing you qualified conversations.
What Changed in v2026.5.5
WhatsApp: Reply-Blocking Stale Clients Stopped Correctly
Two independent fixes address WhatsApp reply delays. Previously, stale local TUI clients that had degraded the Gateway event loop could hold up outbound WhatsApp messages — even when the Gateway itself was healthy. Now only verified stale clients are stopped, leaving healthy Gateway traffic unaffected.
The related /new//reset session-memory hook now runs off the reply path entirely, so memory captures and LLM-slug generation no longer block message delivery. Combined, these two fixes recover a measurable share of response-rate for SDR agents handling 500+ leads per day on WhatsApp.
Feishu: Multi-Turn Topic Sessions Restored
The Feishu channel now correctly hydrates missing native topic starter thread IDs before routing sessions. This means the first turn and all follow-ups in a Feishu conversation stay in the same topic thread — previously they scattered across disconnected sessions, silently fragmenting multi-turn qualification sequences. Fixes #78262.
Feishu is the dominant enterprise messaging platform for Chinese and Southeast Asian manufacturers and exporters — exactly the leads B2B SDR teams target in those markets. Broken topic threading was the kind of failure that's hard to notice until you audit why your Feishu qualification rate is lower than LINE or WhatsApp.
Discord: Live Reasoning + Control Command Routing
Discord users now see live reasoning text in progress drafts instead of a bare Reasoning status line, making agent thinking transparent during complex SDR qualification questions. Separately, plain-text control commands like /steer are now routed through the proper authorization gate instead of being silently dropped before an agent session can process them. Fixes #78080.
LINE: Silent DM Block Prevention
The LINE channel now rejects dmPolicy: "open" configurations that lack a wildcard allowFrom at validation time, instead of silently acknowledging and then blocking inbound DMs. Misconfigurations now surface before leads are lost. Fixes #78316.
Matrix: Approval Delivery Retries
Approval prompts in Matrix now retry up to 3 times with backoff when transient send failures occur. Approval-gated SDR actions — sending a quote, booking a demo, requesting a sample — no longer go missing when Matrix has a momentary hiccup.
iOS Pairing: LAN + Tailscale Flexibility
iOS device pairing now allows setup-code and manual ws:// connects for private LAN and .local gateways while keeping public/Tailscale routes on wss://. Explicit gateway passwords are preferred over stale bootstrap tokens in mixed-auth reconnects. Fixes #47887.
Gateway & Streaming
First-Token Streaming Fixed
The OpenAI-compatible /v1/chat/completions endpoint now flushes the initial assistant-role SSE chunk correctly. Previously, cold agent startup could leave clients with a bodyless 200 response until idle timeout fired — a particularly painful failure for real-time streaming SDR interfaces.
Real-world impact: Custom frontends, Zapier hooks, and CRM integrations built on OpenClaw's OpenAI-compatible API receive the first token immediately instead of hanging.
Gateway Shutdown & Restart Cleanup
Delayed post-ready maintenance is now cancelled during close, and maintenance/cron startup is suppressed after quick restarts. This eliminates a class of orphaned background timers that caused unexpected behavior after Gateway restarts in containerized SDR deployments.
Media Handler Optimization
Media sidecar handling is now skipped for unrelated HTTP routes, so non-media requests no longer pay the media route initialization cost. Gateway response latency under load is reduced.
Control UI: Sessions Table Gets Runtime Visibility
The Sessions table and openclaw status/openclaw sessions CLI output now include an agent runtime column, with filtering by runtime label. SDR supervisors managing mixed Codex/Pi/custom harness deployments can instantly see which runtime handled each session — no more digging through logs to diagnose behavioral differences.
Session history cards have been redesigned with modern checkpoint history UI and a compact N Checkpoint(s) disclosure, reducing visual noise for supervisors reviewing long-running SDR threads.
Provider Fixes: xAI Grok + Fireworks Kimi
Two provider fixes unblock international SDR deployments:
- xAI/Grok: OpenAI-style reasoning effort controls are no longer sent to native Grok Responses models; the bundled xAI thinking profile is clamped to
off. Livexai/grok-4.3runs that were failing withInvalid reasoning effortnow work. - Fireworks/Kimi: K2.5/K2.6 models are exposed as thinking-off-only with
thinking: disabled, so manual model switches no longer send unsupportedreasoning*parameters to Fireworks.
Docker Security Hardening
The bundled docker-compose.yml now drops NET_RAW and NET_ADMIN capabilities and enables no-new-privileges for the gateway container. This is a meaningful security improvement for production B2B SDR deployments running in shared container infrastructure.
Plugin System: Peer Links + Managed Sync
Three plugin fixes reduce operational burden:
- Peer link repair: After any shared-root npm install/update/uninstall, managed plugin
openclawpeer links are re-asserted. One plugin mutation can no longer break SDK resolution for others. - Managed sync during updates: Official plugins (Codex, Discord, WhatsApp, diagnostics) stay synced during host updates even when disabled or previously exact-pinned.
- Corrupt record tolerance: Updates now tolerate corrupt managed plugin records so core package updates complete even when a plugin's metadata is broken.
Session Memory: Collision Fix for /new & /reset
Fallback memory filenames now get collision suffixes when multiple /new or /reset commands fire in the same minute. The earlier session archive is no longer silently overwritten — preserving the full conversation history for SDR audit trails.
How to Upgrade
curl -fsSL https://raw.githubusercontent.com/iPythoning/b2b-sdr-agent-template/main/install.sh | bash
v2026.5.4 vs v2026.5.5: What Changed
| Area | v2026.5.4 | v2026.5.5 |
|---|---|---|
| WhatsApp reply delays | Stale TUI could block any reply | Only verified stale clients stopped |
| Feishu topic threading | First turn splits from follow-ups | Full topic session hydration |
| Discord control commands | /steer silently dropped |
Proper auth + routing |
| Gateway first-token stream | Possible bodyless 200 hang | Immediate SSE flush |
| Docker security | NET_RAW/NET_ADMIN enabled | Dropped + no-new-privileges |
| Plugin peer links | Could break after installs | Re-asserted after every mutation |
| xAI/Grok reasoning | Invalid reasoning effort errors |
Clamped to off |
| Session memory collisions | Same-minute archives overwritten | Collision suffixes added |
Is Your SDR Stack Ready for Production?
v2026.5.5 closes the last remaining reliability gaps for multi-channel B2B SDR deployments. If you're running sales agents on WhatsApp, Discord, Feishu, or LINE, this is a low-risk, high-value upgrade.
Start free → pulseagent.io/app
Related solutions:
- WhatsApp Sales Automation
- AI SDR for B2B Export
- Telegram Lead Generation
- Multi-Channel Sales Pipeline
- AI Sales Agent for Manufacturing
FAQ
Q: Does v2026.5.5 have breaking changes? No. This is a pure patch release — all fixes are backwards-compatible.
Q: Should I upgrade immediately? Yes, especially if you use Feishu, Discord, LINE, WhatsApp, or the OpenAI-compatible gateway. The fixes are targeted and low-risk.
Q: Is the Docker compose change mandatory?
The docker-compose.yml change is included in the bundled template. For existing deployments, adding cap_drop: [NET_RAW, NET_ADMIN] and security_opt: ["no-new-privileges:true"] to your gateway service is recommended.
Q: What exactly broke with Grok 4.3?
xAI's native Grok Responses API does not accept OpenAI-style reasoning_effort controls. OpenClaw was forwarding them, causing Invalid reasoning effort errors on every request. v2026.5.5 strips those parameters for native Grok models.
Q: What's the Feishu topic threading fix? Before this fix, if a Feishu conversation's topic starter thread ID was missing (a transient condition), each turn would open a new standalone session instead of continuing the thread. The fix hydrates the missing ID before routing, keeping the entire conversation in one topic.
