Questions, answered.

Under 10ms per call. The firewall and masking run inline at the edge; budget checks and routing are lookups, not model calls. You will not notice it in a chat UI, and you will see it recorded in the audit log.

One line. Point your OpenAI-compatible client at the Konsole base URL and use a Konsole key. Python, Node.js, LangChain, MCP and plain HTTP all work unchanged. Nothing else in your stack moves.

You bring your own provider keys and pay the provider directly at their rates. Konsole charges for the harness, not for tokens — zero markup. Routing, caching and context discipline then cut that provider bill, typically by half or more.

Names, card numbers, national IDs, health data, API keys and secrets are substituted with placeholders before the prompt leaves your boundary. The model works on the placeholders; the real values are restored in the response on the way back. Your app never knows. Custom regex entities cover anything specific to you.

Our secure cloud, an on-prem box, or inside your own data centre or cloud account. Policies and logs are yours either way.

Every call is logged, attributed to a user, team and app, and exportable to your SIEM. Masking maps to GDPR and HIPAA requirements; logs are immutable for 30 days on Growth and 90 on Business. We do not train on your data. For the current list of formal attestations, ask our security team.

Any provider with an API — GPT, Claude, Gemini, Llama, Mistral, and open-weight models you host yourself. One policy applies to all of them, and you can switch a model out without touching the app.

Yes. Segment sandboxes by department with RBAC, give each team its own budget and model access, and read it all in one audit dashboard.