For organizations that want to keep company data within their own systems and have more control over how AI is deployed, Mozilla is offering an alternative to externally hosted AI services with Thunderbolt, an open-source AI client designed for self-hosted use.

Thunderbolt is available through a waitlist, with native applications for web, macOS, Windows, Linux, iOS, and Android. Its source code is published on GitHub.

Mozilla says Thunderbolt “gives organizations what proprietary AI services can’t,” including ownership of their data, freedom from vendor dependencies, and AI infrastructure they can manage on their own systems with a self-hosted, customizable setup built on open standards.

“Organizations need solutions that provide flexibility and capability without sacrificing control, transparency, or long-term independence,” the company noted.

Through integration with the Haystack platform from deepset, Thunderbolt connects the client interface with backend systems for agent workflows and retrieval based AI applications within a single architecture.

“Integrating with Haystack is a natural opportunity to extend Thunderbolt’s sovereignty to the infrastructure beneath. This gives organizations control not just over how they interact with AI, but how it is built and run,” said Ryan Sipes, CEO of MZLA Technologies Corporation.

In the company announcement, Thunderbolt is described as a sovereign AI client, an open-source, extensible workspace that lets users work with AI through chat, search, and research tools, connect to internal data, and choose the models and tools that suit their needs.

Thunderbolt allows organizations to run AI using a range of models, such as commercial, open-source, and locally hosted options, while connecting to internal systems and data through pipelines and open protocols such as the Haystack platform from deepset, Model Context Protocol servers, and Agent Client Protocol agents.

It supports automation of workflows and recurring tasks, with features like generating briefings, monitoring topics, compiling reports, and triggering actions based on events or schedules. Security features include self hosted deployment, optional end-to-end encryption, and device level access controls.

For enterprise deployments, pricing depends on support, customization, and deployment needs. Integration partners can also provide services such as storage, infrastructure management, and engineering support.

“AI is too important to outsource. With Thunderbolt, we’re giving organizations a sovereign AI client that allows them to decide how AI fits into their workflows – on their infrastructure, with their data, and on their terms,” added Sipes.