Stability & Governance
OpenClaw 2026.5.31 is a milestone for platform stability, introducing the Skill Workshop for better agent capability management and hardening every major communication channel.
Introducing the Skill Workshop
The new Skill Workshop provides a professional environment for developing and deploying agent capabilities.
- Governed Creation: A structured flow for skill proposals, review states, and approval policies.
- Control UI Dashboard: A searchable dashboard for tracking pending revisions, file previews, and deployment history.
- Rollback Safeguards: Automatic metadata tracking for hashes and scanner results ensures you can always revert to a stable state.
- Skill Workshop Tool: A dedicated agent tool allows your agents to assist in the review and application of new skill proposals.
Enhanced Reliability & Recovery
We've implemented significant hardening across all core runtimes to ensure your agents stay responsive.
- Interruption Recovery: Cleaner recovery from interrupted tool calls, stale session bindings, and compaction handoffs.
- Steadier Channels: Mobile delivery is more resilient across Telegram, WhatsApp, iMessage, and Discord.
- Bounded Requests: Better timeout management for provider requests, OAuth device-code lifetimes, and media downloads prevents hangs.
- SQLite State: Persistent state for plugin indexes and inbound queues reduces duplicate filesystem scanning.
Tailscale & Secure Networking
Modern networking support makes it easier to access your OpenClaw instance securely.
- Tailscale Serve: Direct support for Tailscale service-name bindings in the Gateway.
- Service Probes: Hardened local service probes and readiness checks for more accurate health reporting.
- Communication Settings: A new settings tab in the Control UI for granular notification control.
Ecosystem & Model Updates
Support for the latest advancements in the AI provider landscape.
- MiniMax M3: Native support for the MiniMax M3 model family.
- Provider Polish: Metadata fixes for Vertex AI, OpenRouter model caching, and Copilot Claude 1M capabilities.
- External Plugins: Tokenjuice and the GitHub Copilot runtime are now externalized for faster updates.
- Control UI: Improved chat composer controls and agent selectors for Dreaming tasks.
Technical Highlights
- Namespace Support: New internal namespaces for scoped agent and global sessions.
- SQLite Migrations: Backward-compatible migrations for legacy cron and run-log tables.
- Bounded Proof: Diagnostics now cap log sizes to ensure quick reporting of errors.
Upgrade Guide
Who should upgrade?
- Developers creating custom skills via the new Workshop
- Users needing secure remote access via Tailscale
- Anyone experiencing transient mobile notification or channel issues
How to Upgrade
Update your installation via the terminal:
openclaw update
FAQ
What is the Skill Workshop?
The Skill Workshop is a new governed flow for creating, reviewing, and managing skills. it includes a dedicated Control UI dashboard, proposal tracking, and automated rollback metadata.
How does the Tailscale integration work?
OpenClaw now supports Tailscale Serve service-name bindings directly in the Gateway, allowing for easier and more secure remote access with automatic DNS naming.
What reliability improvements are included?
Runtimes now recover more cleanly from interrupted tool calls, and delivery is steadier across mobile channels like Telegram, WhatsApp, and iMessage.
Which new models were added?
This release adds support for MiniMax M3 and improves model metadata for Vertex AI, OpenRouter, and GitHub Copilot Claude 1M capabilities.
Need help from people who already use this stuff?
Join the OpenClaw Community
Get help with your setup, share your agent workflows, and stay up to date with the latest releases in our profit-lab community.