Dropbox Inc. has unveiled new artificial intelligence capabilities designed to simplify how users search for files stored on its platform.
The company also announced its acquisition of Mobius Labs GmbH, a venture-backed developer of AI models. Dropbox plans to leverage the startup’s technology to enhance its platform’s search functionality.
Both announcements center around Dash, a tool first introduced by the company in 2023 that enables employees to search across Dropbox and third-party services like Jira. Until now, Dash was available only as a standalone application.
Dropbox is now rolling out a version of the tool that is fully integrated into its file-sharing platform. Accessible via a new sidebar in the platform’s interface, Dash allows users to input natural language queries—such as “last week’s marketing plan”—instead of exact filenames.
Once Dash locates the requested document, it can summarize its contents or extract specific data points. If the initial response doesn’t fully address a user’s question, they can submit follow-up prompts. Dash can also synthesize information scattered across multiple files into a unified overview.
These new AI features are being released alongside updates to the original standalone Dash application. Previously, customers had to contact Dropbox’s sales team to install the software. Now, employees can bypass that step entirely—a change Dropbox says reduces the setup workflow to just minutes.
The newly launched Dash MCP Server will enable third-party AI applications to interact with Dash. For instance, users of the Cursor coding assistant can now access Dash directly from their interface to search for code files stored in Dropbox without switching contexts. MCP is an open-source protocol designed to facilitate interactions between AI models and external applications.
Looking ahead, Dropbox intends to use technology from its newly acquired startup, Mobius Labs, to further expand Dash’s search capabilities. The startup has developed a suite of multimodal AI models called Aana, which allow users to locate specific moments within videos using natural language prompts. For example, a developer could ask Aana to jump to the segment of a video that explains how to use the newly launched MCP Server.
Aana performs multimodal searches in a hardware-efficient manner through a software module called HQQ. This module handles quantization—a technique used to compress AI models and reduce their memory footprint. According to Mobius Labs, HQQ can quantize models faster than competing technologies.
“Bringing Aana’s capabilities into Dash will not only help us make visual and audio content more searchable,” said Appu Shaji, co-founder of Mobius Labs, in a blog post today. “It will lay the foundation for agent workflows that can analyze and interpret multimedia data, automatically generate insights, and even act on behalf of teams.”
