Big cheeses at the Mozzarella Foundation have released technology which will allow the company’s AI to remain locked inside your own walls.
Mozilla subsidiary, MZLA Technologies, says its Thunderbolt software is an open-source enterprise AI client. It is pitched as a client layer that ties together existing AI models, internal systems, and automation tools into a single interface.
Thunderbolt is framed as an extensible AI workspace where staff use chat, search and research tools to work with AI. The client plugs into enterprise data sources and lets organisations pick what sits behind the interface, whether that is commercial services, open-source models or locally hosted systems.
It ships with native apps for Windows, macOS, Linux, iOS and Android, so staff get the same setup on desktop and mobile.
MZLA Technologies Corporation, CEO Ryan Sipes said: “AI is too important to outsource. With Thunderbolt, we’re giving organisations 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.”
The pitch leans hard on open protocols and familiar plumbing, with support for deepset’s Haystack, Model Context Protocol servers and Agent Client Protocol agents. That means orgs can hook into internal pipelines and automate work without ripping up what they already run.
The Haystack integration is meant to connect the user interface to backend orchestration handling models, retrieval and automated responses.
“Integrating with Haystack is a natural opportunity to extend Thunderbolt’s sovereignty to the infrastructure beneath. This gives organisations control not just over how they interact with AI, but how it is built and run,” Sipes said.
Automation is aimed at scheduled, repeatable work that usually needs someone to keep an eye on it. Thunderbolt can be configured to generate daily briefings, monitor chosen topics, compile reports or trigger actions when incoming data matches a rule.
Security is handled through self-hosted deployment options and encryption settings, along with device-level access restrictions, to reduce reliance on external cloud services with lax data policies. Folding the interface and backend into a single setup gives admins more control over how data moves and how automated tasks execute.
Anyone wanting to test it can join the waitlist at thunderbolt.io, while developers can grab the source from GitHub and start poking around.
