Agent DailyAgent Daily
releaseadvanced

[Release] openclaw/openclaw v2026.5.4: openclaw 2026.5.4

By steipetegithub
View original on github

OpenClaw v2026.5.4 is a performance and stability release featuring improved Google Meet/Voice Call integration with Twilio's realtime Gemini voice bridge, enhanced plugin metadata caching to reduce cold scans, better Windows gateway binding, and numerous UI/UX improvements in the control dashboard. The release includes dependency updates, diagnostic enhancements, and fixes for plugin resolution, secrets management, and channel configuration. Key improvements focus on reducing startup latency, improving voice agent responsiveness, and streamlining operator workflows.

Key Points

  • Google Meet/Voice Call: Twilio dial-in joins now use realtime Gemini voice bridge with paced audio streaming, backpressure-aware buffering, and barge-in queue clearing for snappier voice agent responses
  • Performance: Plugin metadata snapshots are now reused across workspace scopes to eliminate repeated cold plugin metadata scans on hot control-plane paths
  • Gateway/Windows: Default loopback listener now binds only to 127.0.0.1 to prevent libuv dual-stack IPv6 behavior from blocking localhost HTTP requests
  • Plugin auto-enable: Claiming plugin manifest IDs now take precedence over built-in channel aliases (e.g., WeCom/Yuanbao) for proper plugin resolution
  • Secrets management: Auth-profile keyRef and tokenRef fields are preserved during secrets scrubbing to maintain canonical SecretRef metadata
  • Channel resolution: Session-store entries containing ':' are skipped when resolving recall subagent channels to prevent QQ c2c agent ID crashes
  • External channel contracts: npm-published externalized channel plugins now properly resolve compiled artifacts from dist/ directories
  • Models auth: New `openclaw models auth list` command allows inspection of per-agent auth profiles without exposing secrets
  • Control UI: Collapsible New Job sidebar, agent-first chat session filtering, responsive chat controls across devices, and long animation frame diagnostics
  • Gateway startup: Model-catalog helpers, run-session lookup, and schema construction moved out of hot import paths to reduce plugin-load and memory pressure

Found this useful? Add it to a playbook for a step-by-step implementation guide.

Workflow Diagram

Start Process
Step A
Step B
Step C
Complete
Quality

Concepts

Artifacts (4)

models-auth-list-commandcommand
openclaw models auth list [--provider <id>] [--json]
slack-streaming-progress-configyamlconfig
streaming:
  progress:
    render: "rich"
agent-tool-progress-configyamlconfig
agents:
  defaults:
    toolProgressDetail: "raw"
slack-desktop-smoke-testcommand
pnpm openclaw qa mantis slack-desktop-smoke