We're excited to share a round up of the latest features in OpenHands! The latest release of the OpenHands Local GUI brings significant improvements to LLM profile management, enhanced security, and a variety of quality-of-life improvements across the platform. We have also made a few important enhancements and fixes to OpenHands Enterprise.

LLM Profile Management

Note: The following updates will be made available in OpenHands Cloud in a future release.

Saved LLM Profiles — Users can now save and manage multiple LLM configurations as reusable profiles. This eliminates the need to reconfigure model settings when switching between different use cases or providers.

/model Slash Command — A new /model command in chat lets you quickly list and switch between your saved LLM profiles. This provides a seamless way to change models mid-conversation without leaving the chat interface.

Inline Profile Switcher — A convenient profile switcher button is now available directly in the chat input area. One-click model switching means faster workflows for users who frequently alternate between different LLM configurations.

Full Model Name in Header — The conversation header now displays the complete model name instead of abbreviated versions. This clarity helps users quickly verify which model is powering their current conversation.

Sub-agents, Critic, and Sandbox Grouping

Sub-Agent Delegation — OpenHands can now delegate specialized tasks to sub-agents via TaskToolSet. This enables more sophisticated multi-agent workflows where different agents handle different aspects of complex tasks.

Critic Result Display — A new UI component renders critic/verification results inline within conversations. Users can now see quality assessments and validation feedback directly where it's most relevant.

Sandbox Grouping Strategy UI — The sandbox grouping strategy setting is now exposed to all users in the application settings. This gives users more control over how their development environments are organized and managed.

Canonical Settings Query Keys — Frontend settings now use consistent, canonical query keys throughout. This improves reliability and makes debugging settings-related issues significantly easier.

Enterprise Enhancements and Fixes

Tavily MCP Proxy — MCP connections to Tavily are now proxied through the app server for enterprise deployments. This provides better security controls and monitoring capabilities for organizations using Tavily integrations.

LiteLLM Security Update (GHSA-xqmj-j6mv-4862) — Updated LiteLLM to version 1.83.14 to address a security vulnerability. Keeping dependencies current is critical for maintaining a secure software supply chain.

lxml Security Update (CVE-2026-41066) — Updated lxml to version 6.1.0 to patch a known vulnerability. This ensures XML processing remains secure across all document handling operations.

SAML IdP Token Handling — Fixed token storage behavior for SAML identity providers that don't support OAuth tokens. Enterprise users authenticating via SAML will no longer encounter token-related errors.

Organization Settings Fix — Fixed handling of legacy agent_kind='openhands' values in organization responses. Organizations with older configurations will now load correctly without errors.