Bridge is a universal local inference substrate — a way to run capable AI models entirely on your own hardware, on the edge, without sending data to a third-party cloud. It is the foundation the rest of the DVAI ecosystem stands on.
Most AI products send your data to someone else’s cloud to be processed. Bridge inverts that. It runs the model where the data already lives — on the device, on-premise, inside your own network — so sensitive information never has to leave the building.
For regulated manufacturers, healthcare providers and anyone whose data is their competitive edge, that distinction is the difference between adopting AI and not being able to. Bridge launched on 15 May 2026, with a UK provisional patent filed.
Bridge is a universal local-inference substrate — the layer that lets capable AI run directly on a device or inside your own infrastructure, in-process, rather than calling out to a third-party cloud. It is the sovereignty foundation the rest of the DVAI ecosystem is built on: when an agent runs on-device or behind the firewall, Bridge is what makes that possible.
Bridge is source-available under a BSL-style licence — the code is visible and usable, with commercial terms attached, rather than open-source in the strict sense. It is free for personal, educational, academic, evaluation and internal-only use. Commercial use carries a fair, openly-published tiered royalty: the first £90,000 of revenue is free, then 4% / 6% / 8% on revenue bands. Every release automatically converts to a permissive Apache 2.0 licence three years after it ships.
Bridge presents ten inference backends behind a single OpenAI-compatible API — llama.cpp, MediaPipe, MLX, Transformers.js, WebLLM, LiteRT, LiteRT-LM, ONNX Runtime, ML.NET and llamafile — and selects the right engine per device. It ships as SDKs for TypeScript, Swift, Kotlin, Dart and C#, with React Native and Capacitor wrappers, so you work in the language your stack already uses.
Bridge is a library, not a separate product — it runs inside your app’s own process, with no sidecar, no Docker and no separate server for your users to install. Inference happens locally, so the data being processed does not leave the device or your environment to reach a model. There is no cloud round-trip in the inference path.
They can pair. A phone can offload heavier inference to a paired laptop on the same network, end-to-end encrypted, so you get more capable local inference without sending data to a public cloud. Each device can also run entirely on its own where that is what you need.
Yes — Bridge is patent pending. A UK patent application (GB2611312.6, filed 14 May 2026, applicant Deep Voice AI Limited) covers its core approach, its novelty features and its uniqueness.
Bridge is live now. Evaluation can start with the documentation and an internal-evaluation licence, which is free, then a scoped trial in your own environment. Because it runs locally, you can test it against your own data-residency and performance requirements directly, rather than taking our word for it.
Bridge is the sovereignty layer beneath the ecosystem. Request a confidential technical conversation with our team.
Every enquiry is held in confidence. We’ll walk you through the use cases relevant to you and scope how a sovereign deployment would work in your environment — with no obligation to proceed.
Capable AI, on your terms — not the cloud’s.