thunderbolt

Thunderbolt is an open-source AI chat client from the team behind the Thunderbird email application. Its core idea is that you should be able to pick which AI models you talk to, keep your data under your own control, and avoid being locked into any single AI provider. It is designed mainly for companies that want to run their own AI setup on their own servers rather than relying on cloud services.

The app runs on web, Mac, Windows, Linux, iOS, and Android. You connect it to AI models of your choosing: local models running on your own machine via tools like Ollama or llama.cpp, private on-premises servers, or any cloud AI provider that is compatible with the OpenAI API format. You add your own API keys or point it at your own server; there is no built-in inference endpoint provided by the project itself.

To get it running locally, the README describes a short sequence of make commands: one to verify your tools are installed, one to install dependencies, one to start a database and a sync service in Docker, and one to start the backend and frontend. For running it in production, the project provides Docker Compose and Kubernetes configuration files. The README notes that it currently requires internet access for authentication and optional search features, though both can be configured or disabled.

The project is still in early development and was undergoing a security audit at the time of writing. The README explicitly cautions that it is not yet production-ready and that the team is prioritizing enterprise customers who want on-premises deployment. Community contributions are welcome, and bugs or feature requests go through GitHub issues.

Thunderbolt is licensed under the Mozilla Public License 2.0.