[Release] openclaw/openclaw v2026.5.19-alpha.1: openclaw 2026.5.19-alpha.1

OpenClaw v2026.5.19-alpha.1 introduces significant enhancements across agents, dependencies, Docker/Podman support, gateway performance, plugins, skills, browser automation, QA testing, and platform-specific features. Key updates include refactored agent fixes with clean bounded refactors, new skill capabilities (meme-maker, Python debugging, node inspector), improved Settings UI on Mac, enhanced browser dialog handling, and comprehensive QA-Lab runtime parity scenarios. The release also addresses critical fixes for port validation, plugin tool auth profiles, and memory search performance optimization.

Key Points

• •Update dependencies: @openclaw/proxyline to 0.3.3, Pi packages to 0.75.1, and raise minimum Node.js to 22.19
• •Add runtime-neutral Docker/Podman build args (OPENCLAW_IMAGE_APT_PACKAGES, OPENCLAW_IMAGE_PIP_PACKAGES) for flexible image customization
• •Implement Gateway startup probe optimization to reduce restart ready latency while preserving readiness gating
• •Introduce new skills: meme-maker (template search, SVG/PNG rendering, Imgflip support), Python debugging (pdb, breakpoint, debugpy), and node inspector with spike workflows
• •Enhance browser automation with modal dialog tracking, blockedByDialog responses, and configurable timeout-ms for long-running page functions
• •Add CLI plugin tooling: defineToolPlugin, openclaw plugins build/validate/init for typed simple tool plugins with generated manifests
• •Expand QA-Lab with runtime parity scenarios (20-turn, 100-turn), tool fixture coverage reporting, and live-frontier token-efficiency testing
• •Redesign Mac app Settings with consistent card layouts, cached navigation, and refined Voice & Talk recognition settings
• •Support HTTPS managed forward-proxy endpoints with scoped TLS CA trust configuration
• •Fix critical issues: port validation (reject >65535), plugin tool auth profile preservation, and memory search performance with bounded rowid batches

Workflow Diagram

Concepts

Artifacts (5)

# Surface pending and recently handled modal dialogs
openclaw browser dialog --dialog-id <id>

# Evaluate with extended timeout
openclaw browser evaluate --timeout-ms 30000 <script>

# QA-Lab runtime parity tier configuration
openclaw qa suite --runtime-parity-tier standard

# Tool fixture coverage reporting
openclaw qa coverage --tools

# Scenario tiers:
# - standard: 20-turn Codex-vs-Pi parity
# - optional: 100-turn extended scenarios
# - live-only: frontier canary and self-health tests
# - soak: long-running stability tests

# Runtime-neutral image build arguments
OPENCLAW_IMAGE_APT_PACKAGES=<extra-apt-packages>
OPENCLAW_IMAGE_PIP_PACKAGES=<python-packages>

# Legacy fallback
OPENCLAW_DOCKER_APT_PACKAGES=<extra-apt-packages>

import { defineToolPlugin } from '@openclaw/plugins';

export default defineToolPlugin({
  name: 'my-tool-plugin',
  tools: [
    {
      name: 'my-tool',
      description: 'Tool description',
      inputSchema: {
        type: 'object',
        properties: {
          param1: { type: 'string' }
        }
      },
      execute: async (input) => {
        // Tool implementation
        return { result: 'output' };
      }
    }
  ]
});

# Benchmark repeated restart readiness, downtime, trace, and resource-slope
pnpm test:restart:gateway

# Measures:
# - Startup probe attribution
# - Config and runtime costs
# - Resource-count costs in restart traces
# - Overlap startup logging with plugin-service startup