With this release, we bring you a function to run Agents with Python SDK, SuperAnnotate MCP servers for your agents. claude-opus-4-5 and gpt-5.2 support, and much more!

Agent hub updates

SuperAnnotate MCP Server

The SuperAnnotate MCP Server provides access to SuperAnnotate tools for seamlessly managing projects, folders, and items through your agent in the Agent Hub.

Run Agents function

You may now use the runAgents() function to trigger one or multiple agents, each defined by a configuration object specifying the model, prompt, parameters, and expected response format.

New Anthropic model

You may now select the claude-opus-4-5 model for your Anthropic agents.

New Openai model

You may now select thegpt-5.2 model for your OpenAI agents.

Multimodal updates

• You may use the Run Model component, its inputs and its outputs with various Grouping structures.
• The on_componentID_tab_change event has been added to Tabs components.
• The setOptions() function for Dynamic Select components now works for multi-select as well.

Explore updates

• We’ve added an Explore variant for CSV downloads. You may now select Explore-specific columns that you want to export and download their data in CSV format.
• Input values are no longer mandatory for the Run Model feature in Explore.
• When viewing Multimodal items in Zoom View, you may hover over each component to see its component type and ID, which can be easily copied if needed.

Additional features & improvements

• Using any of the following special characters in your item names will now cause the item to be skipped entirely when imported. The special characters are: < > : ” ' / \ | ? * & $ ! +