Agent DailyAgent Daily
releaseintermediate

[Release] openclaw/openclaw v2026.5.19: openclaw 2026.5.19

By github-actions[bot]github
View original on github

OpenClaw v2026.5.19 is a major release focusing on agent architecture improvements, dependency updates, and enhanced QA testing capabilities. Key updates include refined plugin SDK deprecation paths, Docker/Podman build argument standardization, Gateway startup optimization, new skills (meme-maker, Python debugging, node inspector), and comprehensive runtime parity testing through QA-Lab. The release also improves Mac app UI consistency, browser modal dialog handling, and adds typed tool plugin support via `defineToolPlugin`.

Key Points

  • Agent fixes should default to clean bounded refactors with lean internals and explicit plugin SDK/API deprecation paths
  • Updated dependencies: @openclaw/proxyline to 0.3.3, Pi packages to 0.75.1, minimum Node.js raised to 22.19
  • Docker/Podman: added `OPENCLAW_IMAGE_APT_PACKAGES` and `OPENCLAW_IMAGE_PIP_PACKAGES` for flexible runtime-neutral image builds
  • Gateway startup optimized: overlapped plugin-service startup with channel sidecars to reduce restart ready latency while preserving `/readyz` sidecar gating
  • New skills added: meme-maker (template search, SVG/PNG rendering, Imgflip support), Python debugging (pdb, breakpoint, debugpy), node inspector, and autoreview
  • CLI enhancements: `openclaw skills install/update --global` for shared managed skills, `openclaw browser evaluate --timeout-ms` for long-running functions, `defineToolPlugin` for typed tool plugins
  • QA-Lab expansion: first-hour 20-turn and 100-turn runtime parity scenarios, live-only canaries, tool fixture coverage reporting, and hard-gated dynamic runtime-tool drift verification
  • Android Talk Mode switched to realtime Gateway relay with streaming mic input, realtime audio playback, and tool-result bridging
  • Browser improvements: surface pending/recently handled modal dialogs in snapshots, return `blockedByDialog` when action opens modal, allow dialog answering via `browser dialog --dialog-id`
  • Mac app Settings redesigned with consistent card layouts, cached navigation, and cleaner permissions/voice/skills/cron/exec/debug panes

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)

Docker Build Argumentsbashconfig
OPENCLAW_IMAGE_APT_PACKAGES=<apt-packages>
OPENCLAW_IMAGE_PIP_PACKAGES=<python-packages>
OPENCLAW_DOCKER_APT_PACKAGES=<legacy-fallback>
CLI Tool Plugin Definitionbashcommand
openclaw plugins build
openclaw plugins validate
openclaw plugins init
openclaw skills install --global
openclaw skills update --global
openclaw browser evaluate --timeout-ms <ms>
openclaw qa suite --runtime-parity-tier
openclaw qa coverage --tools
openclaw codex plugins list
openclaw codex plugins enable <plugin>
openclaw codex plugins disable <plugin>
Gateway Startup Configurationyamlconfig
# Gateway startup probe configuration
# - Attribute startup probe, config, runtime, and resource-count costs in restart traces
# - Overlap startup logging and plugin-service startup with channel sidecars
# - Reduce restart ready latency while preserving /readyz sidecar gating
gateway:
  startup:
    overlappedPluginStartup: true
    channelSidecarConcurrency: true
    readinessGating: /readyz
QA-Lab Runtime Parity Configurationyamlconfig
qa:
  suites:
    runtimeParity:
      tiers:
        standard:
          turns: 20
          gating: hard
          verification: codex-vs-pi
        extended:
          turns: 100
          gating: optional
      scenarios:
        - first-hour-20-turn
        - first-hour-100-turn
        - live-only-canary
        - tool-fixture-coverage
        - personal-agent-approval-denial
        - personal-agent-task-followthrough
        - personal-agent-dreaming-shadow-trial
        - personal-agent-share-safe-diagnostics
        - personal-agent-no-fake-progress