What's new in AIsuru | v11.1.0

Welcome back to the AIsuru release notes v11.1.0! Here you'll find new features, improvements, and bug fixes to make your AI Agents more powerful and secure 🚀

New prompt architecture

We've restructured our agents' system prompts to improve both comprehension and efficiency, in terms of token usage and caching. This results in more consistent responses and more efficient use of resources, with no visible impact on how you configure your agents.

MCPs and groups

We introduced a new "Groups" tab with targeted visibility for corporate MCPs, and improved the MCP section with new actions and a clearer layout.

Agent creation and editing

When creating a new agent, new default values now allow for longer responses, with the maximum response token count initially set to 32,000. The new agent's language is also preselected automatically based on the browsing language.

Web address links, on both private and public layouts, now open in the same window instead of a new one.
Agent management pages also sync the active tab in the URL: this lets you share a direct link to a specific tab or panel with colleagues and other authors, keeps the browser's Back/Forward buttons consistent, and makes sure a page refresh doesn't lose your place.

Authentication and error handling

422 error handling has been reworked: private agents now trigger a password modal (treating the case as a 403 error), while public agents show an explicit, immediately understandable warning.

Loading states and session management

The chat layout's loading indicator is now shown only once the session has actually started. The message input stays disabled while there's no active session (unless the microphone is on), and only receives a session ID after the session starts. In the hidden chat layout, the start panel stays visible until the chat actually replaces it.

Interface and styling

We cleaned up the CSS by removing a large amount of duplicated rules: the "Disabled" status color is now consistent throughout the chat, and z-index calculations follow a single, uniform scheme. The channel access section has been reorganized for better readability, and agent reasoning is now shown in the conversation archive only to authors.

Models and technical integrations

AWS Bedrock now uses base64 encoding for image uploads.
We added support for the Anthropic claude-haiku-5-5 model and updated vulnerable libraries across the platform.

🐞WHAT WE FIXED!

Markdown tables have been adapted to match the chat interface's style.
Fixed an issue where the avatar image was rejected as "too large" right after cropping, even when the original file was under 2 MB.
Fixed a bug where the avatar preview overlapped with text.
The loading indicator inside buttons no longer oscillates off-center.
Fixed UI state handling when removing media files from documents uploaded in chat.

🛠️ As always, we’ve addressed various minor bugs and implemented performance improvements across the platform to ensure a smoother, faster experience.

💡 Need help with AIsuru? Want to explore the AIsuru AI Academy? Contact us at demo@memori.ai