Mozilla has announced the release of Thunderbolt, an open-source client designed to let organizations run and interact with artificial intelligence systems on their own infrastructure, without relying on centralized cloud platforms. The project was made public on April 16, 2026, and is developed by MZLA Technologies, Mozilla’s for-profit subsidiary best known for maintaining the Thunderbird email application.
Thunderbolt is aimed at developers, researchers, and organizations that want more control over how AI is deployed and used. Rather than offering its own models, the software acts as a client layer, allowing users to connect to AI systems they host themselves. This approach reflects growing demand for tools that keep sensitive data inside private networks and reduce dependence on third-party AI providers.
According to Mozilla, Thunderbolt provides a single interface through which users can interact with AI models for tasks such as conversation, research, and information retrieval. System administrators are able to decide which models back the interface, with support for both commercial offerings and open-source alternatives. For teams running models locally, Mozilla points to existing inference engines such as Ollama and llama.cpp, which can be used alongside the client.
Thunderbolt is also built to integrate with existing AI workflows instead of replacing them. The software supports the Model Context Protocol, an open standard intended to connect AI models with internal data sources. It can also be combined with Haystack, an open-source framework developed by deepset for retrieval-augmented generation and orchestration, making it easier for teams to connect AI systems to their own documents and pipelines.
Mozilla emphasizes that Thunderbolt is open source and published under the Mozilla Public License 2.0, with the code available publicly on GitHub. The company describes the project as early-stage but usable for experimentation and internal deployments, particularly for organizations evaluating self-hosted AI setups. Current testing is supported through custom back ends, including environments deployed via Docker.
While Mozilla has signaled possible future commercial offerings—such as professional services or managed deployments—it has not announced pricing or timelines. For now, Thunderbolt is positioned as a foundation: a client that allows organizations to decide how, where, and under what conditions AI fits into their workflows.
As debate grows around data sovereignty, platform concentration, and the long-term costs of cloud-based AI, Mozilla’s Thunderbolt reinforces a familiar stance for the company: favoring open software, local control, and user choice over closed systems.
Sources: