Reliability & model routing
How Syncanix keeps answering: multiple model providers, your own keys, and EU routing.
A useful assistant has to stay up. Syncanix is built so that a single model provider’s bad day doesn’t become yours, and so you keep control over cost and where traffic goes.
More than one model provider
By default Syncanix runs Claude on AWS Bedrock in the EU, and can fail over automatically to the first-party Anthropic API or OpenAI when a provider is unavailable or slow. Your assistant keeps answering through a provider outage.
Bring your own keys
Prefer to use your own provider accounts? With bring-your-own-key, model usage is billed directly to you at your negotiated rates, with no Syncanix markup. You can set this per provider and per environment.
EU routing
Everything Syncanix stores stays within the European Union, and so does default AI inference — Claude runs on Amazon Bedrock in eu-central-1, and retrieval embeddings and reranking run there too. If the EU model path is unavailable we can fail over to a provider outside the EU, and with BYOK you choose the model provider — and that data path — explicitly.
Cost stays in view
You can see what the assistant costs and set spend caps — per workspace, per user, or per conversation — so usage never runs away from you. Usage and cost live under Settings → Billing.