Private, self-hosted inference

Security research can involve unpublished source code, private firmware, and findings that cannot be sent to a hosted model service. Vox is designed for these environments, using a self-hosted OpenAI-compatible inference server.

Your model, your infrastructure

You operate the model endpoint used for the investigation. It can run on your workstation or within your own research infrastructure.

The conversation and relevant research results go to that endpoint for model processing. This gives you control over where the inference happens and how the server handling your research data is operated.

Deliberate access

You choose the material available to agents and the actions they can take. Research environments can restrict network access while keeping the model reachable. Online research is available when the investigation permits it.

Vox has no analytics or telemetry. Saved conversations and research work remain in your Vox environment.

Keep useful automation

Private inference still supports coordinated agents, research tools, and ongoing investigations. You can delegate analysis and inspect the evidence while keeping the model processing within your own infrastructure.

Continue with agent orchestration or research environments.