Thunderbolt is MZLA Technologies' (Mozilla's for-profit subsidiary, makers of Thunderbird) answer to enterprise AI lock-in. Announced April 16, 2026, it's an open-source, cross-platform AI client organizations can deploy on-prem or in their own cloud. No data sent to Microsoft, OpenAI, or Anthropic servers by default. Out of the box, Thunderbolt connects to cloud providers (Anthropic, OpenAI, Mistral, OpenRouter) or local models via Ollama and llama.cpp. Organizations bring their own models, run on their own infrastructure, and own all conversation data. Available on web, iOS, Android, Mac, Linux, and Windows under Mozilla Public License 2.0.
Thunderbolt is Mozilla’s strategic entry into enterprise AI — specifically targeting organizations that have evaluated Microsoft Copilot, ChatGPT Enterprise, and Claude Enterprise and decided that vendor control of their data is a dealbreaker.
Announced April 16, 2026 by MZLA Technologies, the for-profit subsidiary of the Mozilla Foundation and maintainers of the Thunderbird email client, Thunderbolt is positioned as the open-source alternative to closed enterprise AI clients. The core pitch: deploy it yourself, connect whatever models you choose, and your conversation data never touches a third-party server you don’t control.
Model-agnostic backend: Thunderbolt supports Anthropic (Claude), OpenAI (GPT series), Mistral, and OpenRouter as cloud providers out of the box. For organizations running local models, it integrates with Ollama, llama.cpp, and any OpenAI-compatible API endpoint.
Data sovereignty architecture: By default, conversations are stored in your own infrastructure. The backend is wired to your chosen model providers rather than routing through Thunderbolt’s servers. Organizations with strict data residency requirements (healthcare, legal, government) get compliance without sacrificing AI capability.
Cross-platform client: Web, iOS, Android, Mac, Linux, and Windows are all supported — aimed at the enterprise need for consistent experience across device types without separate deployments.
Thunderbolt is early and under active development. It’s best suited for teams with technical capacity to run Docker deployments who are currently paying for enterprise AI plans and are uncomfortable with where their data goes. The Mozilla brand brings credibility to the data ownership claim — MZLA has no revenue model that depends on analyzing your conversations.
Less suited for teams without infrastructure expertise or those who need mature enterprise features (SSO, SCIM, advanced audit logs) — these are on the roadmap but not yet shipped.
Unlike Claude Enterprise or ChatGPT Enterprise, Thunderbolt is MPL 2.0 and self-hostable with no per-seat licensing. Unlike LibreChat (another open-source AI client), Thunderbolt carries the Mozilla institutional backing and is targeting the enterprise market explicitly rather than hobbyist self-hosters.
Opinionated bundle of agent personas — pre-built role templates for engineering, marketing, ops, and research teams that ship as drop-in markdown files.
OpenAI's flagship AI assistant capable of conversation, analysis, coding, writing, and multimodal reasoning across text and images.
Claude Code session memory plugin that captures context, compresses it via Agent SDK, and injects relevant memory into future sessions automatically.