Dropbox Inc. today introduced new artificial intelligence features that will make it easier for users to search the files they keep in its platform.

The company also announced that it has acquired Mobius Labs GmbH, a venture-backed AI model developer. Dropbox will use the startup’s technology to enhance its platform’s search features.

Both announcements focus on a tool called Dash that the company introduced in 2023. It enables workers to search for files across Dropbox and third-party services such as Jira. Until today, Dash was only available as a standalone application.

Dropbox is now rolling out a version of the tool that is integrated into its file sharing platform. Dash is accessible through a new sidebar in the platform’s interface. It enables users to enter natural language descriptions such as “last week’s marketing plan” instead of specific file names.

After Dash finds a document requested by the user, it can summarize its contents or extract specific data points. Workers can enter follow-up requests if the initial prompt response doesn’t fully answer their questions. Dash is also capable of consolidating information spread across multiple files into a single overview.

The new AI features are rolling out to Dropbox alongside an update to the original standalone Dash application. Previously, installing the software required customers to contact the software maker’s sales department. Workers can now skip that step, which Dropbox says reduces the installation workflow to a few minutes.

A newly released MCP server for Dash will enable third-party AI applications to interact with the tool. Users of the Cursor programming assistant, for example, can now access Dash without leaving the interface and have  it search code files stored in Dropbox. MCP is an open-source protocol designed to let AI models interact with external applications. 

Going forward, Dropbox plans to extend Dash’s search capabilities using technology from a newly acquired startup called Mobius Labs. The startup has developed a series of multimodal AI models called Aana. The algorithms enable users to find specific moments in a video with natural language prompts. A developer, for example, could ask Aana to jump to the part of a video that explains how to use a newly launched MCP server.

Aana can carry out multimodal searches in a hardware-efficient manner thanks to a software module called HQQ. The module performs quantization, which is the task of compressing an AI model to reduce its memory requirements. Mobius Labs says HQQ can quantize models faster than competing technologies. 

“Bringing Aana’s capabilities into Dash won’t just help us make visual and audio content more searchable,” Mobius Labs co-founder Appu Shaji wrote in a blog post today. “It’ll provide foundational support for agentic workflows that can analyze and interpret multimedia data, surface insights automatically, and even take action on behalf of teams.”

Image: Dropbox