Mozilla unveils Thunderbolt, an open-source AI client for private, self-hosted workflows

Mozilla has introduced Thunderbolt, an open-source AI client designed to give organizations direct ownership of their data, self-hosted infrastructure, and complete freedom from vendor lock-in. By offering a platform that is customizable and developed on open standards, Thunderbolt places control over AI operations fully in the hands of the user.

Building on this foundation, Thunderbolt enables organizations to run AI workflows entirely on their own infrastructure. Teams can select their preferred machine learning models, connect the system to proprietary data sources, and safeguard privacy across operations. These capabilities are available without reliance on third-party cloud services.

While flexibility and control are at the core, Thunderbolt extends its reach with native integration of deepset’s Haystack framework. This partnership connects user-facing chat, search, and research capabilities with enterprise-grade agent orchestration and retrieval augmented generation (RAG) in a unified architecture. Organizations can thus automate workflows, manage recurring tasks, and streamline research directly within the platform.

Thunderbolt’s open source code is accessible on GitHub under the MPL-2.0 license, and native applications are available for web, macOS, Windows, Linux, iOS, and Android via a waitlist. Mozilla’s collaboration with deepset ensures aligned development between the client interface and backend orchestration layers.