OpenClaw v2026.3.12 shipped on March 13, 2026, with 40 contributors delivering a completely refreshed dashboard, fast mode support for major providers, and over a dozen security fixes.
Dashboard v2
The gateway Control UI gets a full refresh with modular views for overview, chat, config, agent, and session management. A new command palette provides keyboard-driven navigation, while mobile bottom tabs improve the experience on smaller screens. Chat gets richer tooling including slash commands, search, export, and pinned messages.
Fast Mode
A new fast mode toggle is now available for both OpenAI/GPT-5.4 and Anthropic/Claude, with configurable settings. This allows users to trade some response quality for significantly faster agent replies when speed matters more than depth.
Provider Architecture
Ollama, vLLM, and SGLang have been moved to a provider-plugin architecture, making it easier to add and maintain local model providers. Kimi Coding compatibility is restored with native Anthropic-format tool support.
Security Hardening
The release addresses 14+ security vulnerabilities:
- Device pairing: Short-lived bootstrap tokens replace embedded gateway credentials in QR/chat pairing
- Workspace plugins: Implicit auto-load disabled — cloned repos can no longer execute plugin code without explicit trust
- Exec approvals: Invisible Unicode format characters escaped so zero-width text renders as visible escapes
- Webhook signatures: Enforced across multiple platforms
Infrastructure
A new Kubernetes starter path with raw manifests and Kind setup makes it easier to deploy OpenClaw on K8s clusters. For users who prefer managed deployment, OpenClaw Launch handles all infrastructure automatically.