Mozilla, the organisation behind the Firefox browser and long-time champion of open-source software, has moved into the enterprise AI market. On April 16, 2026, MZLA, Mozilla’s for-profit subsidiary, announced Thunderbolt, an open-source AI client that lets businesses run AI workloads entirely on their own infrastructure, without routing sensitive data through third-party cloud services.

The launch positions Thunderbolt as a direct alternative to Microsoft Copilot and other cloud-hosted AI assistants, specifically targeting organisations where data residency, security, and auditability are non-negotiable.

What Thunderbolt Actually Does

At its core, Thunderbolt is an enterprise AI client that connects to models and orchestration systems of your choosing rather than locking you into one provider’s stack. It is built on deepset’s Haystack platform for AI orchestration and infrastructure, and supports the Model Context Protocol (MCP) and the Agent Client Protocol (ACP).

That means businesses can connect Thunderbolt to commercially hosted models like Anthropic’s Claude or OpenAI’s Codex, open-source models running locally, or entirely self-hosted deployments. The same interface, regardless of what is running underneath.

Key features announced at launch include:

• Self-hosted deployment with an offline SQLite database as a local source of truth, keeping your data on your own machines
• Optional end-to-end encryption and device-level access controls for organisations with strict compliance requirements
• Workflow automation for recurring tasks, built on top of MCP and ACP integrations
• Cross-platform native apps for Windows, macOS, Linux, iOS, and Android
• GitHub availability immediately, with a managed hosted version in development for organisations that prefer not to manage infrastructure themselves

Enterprise pricing is flexible, depending on support level, customisation needs, and deployment model. Integration partners also offer storage, infrastructure management, and engineering support services.

Why Mozilla and Why Now

Mozilla’s brand is built on one thing: trust. The organisation spent decades arguing that the web should be open, private, and user-controlled. Applying that same philosophy to enterprise AI is a natural extension, particularly as more businesses grow uncomfortable with the data implications of using large cloud-hosted AI platforms.

The timing is also notable. Microsoft Copilot is now deeply embedded in the Microsoft 365 stack, which means many organisations are defaulting to it simply because it is already there rather than because they have actively evaluated it. Thunderbolt gives procurement teams and IT departments a credible alternative with a clear privacy and control story.

The choice to build on deepset’s Haystack platform is a smart one. Haystack is a well-regarded open-source AI orchestration framework with genuine enterprise adoption, meaning Thunderbolt is not starting from scratch on the harder infrastructure problems.

What This Means for Business

For business owners and technology leaders, Thunderbolt raises a useful question that should already be on your agenda: where is your AI data actually going?

Most off-the-shelf AI tools, including many popular productivity assistants, send your prompts and documents to provider infrastructure for processing. For many use cases that is perfectly acceptable. But for businesses handling sensitive client information, proprietary financial data, regulated healthcare records, or confidential internal planning, the calculus changes.

Thunderbolt is not the only self-hosted option, but Mozilla’s involvement lends it a level of credibility that most open-source alternatives lack. The Mozilla brand signals long-term commitment to the project and a track record of building for maintainability rather than a quick exit.

A few things to watch as the product develops:

Performance parity. Self-hosted and locally run models often lag behind the latest frontier models in raw capability. Thunderbolt’s ability to route to commercial APIs helps here, but organisations will need to manage model selection carefully.

Managed hosting roadmap. The planned managed hosted version will be key for smaller organisations without dedicated infrastructure teams. How Mozilla prices and secures that offering will determine whether Thunderbolt becomes genuinely mainstream or stays a specialist tool for technical buyers.

Integration depth. MCP and ACP support is a solid foundation, but enterprise workflows are complex. The quality of pre-built connectors to common business systems, from CRMs to ERP platforms, will matter.

---

Mozilla entering enterprise AI is not a surprise in retrospect, but it does shift the dynamics of the market. There is now a credible, open-source, privacy-first option for businesses that want AI capability without handing over their data to a hyperscaler.

If you want the playbook other teams are using with Claude and Codex right now, grab the free Working With Claude field guide. Download it here.