Multi-model AI deliberation, on your own servers, under your control.
The engine adds the provider key and calls the provider itself; no browser, script or laptop ever holds a key, and keys never reach the logs.
The engine faces your people; a separate Admin Server keeps accounts, passwords and keys, and it can be reached only from the server itself, never over the network.
Each person signs in with their own account; many sessions run side by side, each in its own process, each with only the models its person may use. Your team works in the browser. Access tokens are there if you also want scripts.
Seats, administrators, per-model access, revocable access tokens for scripts, lockout of repeated failed sign-ins, a complete audit log, and a live view of running sessions.
Checked with a packet capture across restarts and real sessions: zero connections to Osinix. Restarts by itself if stopped; starts at boot.
x86-64 and ARM servers. Runs beside existing systems: anything that speaks https can start sessions.
MCS adds a governed deliberation layer to the servers you already offer: the hardware supplies the computing, MCS supplies multi-model reasoning with enterprise control.
MCS works with models on your own servers, cloud models, or a mix, and can run entirely on models on your own servers.
With cloud models, what people send goes to the provider your administrator chooses. Everything else stays on your server.
