One control layer over your models, data, knowledge and region.
Run open-weight models on dedicated, region-pinned infrastructure through an OpenAI-compatible API — with your data staying exactly where you put it.
The control layer, in three pillars
Everything on the platform answers to the same three decisions: what runs, what it remembers, and where it lives.
Models
You choose which model runs your workload. Qwen 3.8 with a 1M context window runs on dedicated servers; the lineup grows through the roadmap.
- Qwen 3.8 with a 1M context window
- Runs on dedicated servers
Data & knowledge
Prompts and outputs are not stored by default, and are never shared or used to train shared or third-party models. Memory — the knowledge pillar — is private to your organization: encrypted, pinned to your region, persistent, isolated by organization and permissions, under your retention and access policies.
- Prompts & outputs not stored by default
- Encrypted, region-pinned, org-isolated
- Your retention & access policies
Region
EU, USA or LATAM: your workloads run in the region you choose, and your endpoints name it.
The control plane, capability by capability
Every part of the platform below runs under the same identity, guardrails, knowledge and regional endpoint — each page documents what is live today and what is planned.
Your endpoint names your region.
Pick where your data must stay. Every call goes to an endpoint that says exactly where it runs.
Start with control.
Start with control: create your account and pick the region your data must stay in.