Tool Updates

The read file tool now supports PDF, email, and image formats with limit/offset pagination that defaults to the first ten pages while informing the agent about remaining content. For images, the agent can describe visual contents directly. Every new agent ships with read file and read note tools enabled by default, saving tokens and improving speed by skipping unnecessary context injection.

Search now scopes results by space when operating within a task context, and searches across all accessible spaces when used by the assistant agent. The create note tool respects task ID or space ID parameters, defaulting to the user’s My Tasks space when neither is provided. Set task status works across different tasks, not just within the current task context.