![]()
Mozilla subsidiary MZLA announced Thunderbolt, which it describes as an open source AI client for businesses.
“Thunderbolt is an open-source AI client that gives organizations what proprietary AI services can’t: Full ownership of their data, freedom from vendor dependencies, and AI infrastructure that stays entirely within their hands – self-hostable, customizable, and built on open standards,” the announcement post explains. “Through a native integration with deepset’s Haystack, Thunderbolt extends control to the infrastructure layer – connecting the client experience with enterprise-grade agents and RAG [retrieval augmented generation] orchestration within a unified architecture.”
Thunderbolt is a sovereign AI client, MZLA says, an open source and extensible workspace where users can interact with the AI of the company’s choice using chat, search, and research while using internal data. Companies can choose from commercial, open source, and local AI models, integrate with systems and data using Haystack, Model Context Protocol (MCP) servers, and agents with the Agent Client Protocol (ACP), automate workflows and recurring tasks, work seamlessly with apps in Windows, Mac, Linux, Android, and iOS, and secure the entire thing with end-to-end encryption and device-level access controls.
MZLA created Thunderbolt in partnership with Berlin-based Haystack maker Deepest, which has enterprise and public sector customers across government, aerospace, and multinational. And it says the two have created something unique that provides both the underlying architecture and infrastructure and, now, with Thunderbolt, the AI client experience.
“AI is too important to outsource,” MZLA Technologies CEO Ryan Sipes says. “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.”
Thunderbolt is available now, but there’s a waitlist and organizations are advised to contact MZLA directly if they’re interested in a pilot deployment and enterprise licensing. There are clients for Windows, Mac, Linux, iPhone, iPad, and Android, and the source code is available on GitHub.