Sandbox
Audit-chain demo
Open a sandbox tenant. Send a prompt. Watch every model call land in the audit chain — signed, hash-linked, externally verifiable in under a second.
OpenPillar 1 · Control · See · Change · Walk away
You're using AI for real work, but you can't see which models your team is calling, what context is leaving your network, or what happens to it on the other side. If a vendor changes terms, removes a feature, or goes down, your workflows break.
Evidence
No screenshots-only feature pages. Every item below runs against a sandbox, downloads as a real artefact, or links to an externally-verifiable record.
Sandbox
Open a sandbox tenant. Send a prompt. Watch every model call land in the audit chain — signed, hash-linked, externally verifiable in under a second.
OpenInteractive
Inspect every routing decision: which tenant, which user, which sensitivity tag, which policy bundle, which model picked. No black boxes.
OpenA standard MSA clause that ships in every Operayde contract. Annual cancellation right with full export of corpus + weights + LoRA + audit chain. Operayde retains nothing.
OpenInteractive
Pin a specific model version. See diff against the previous one when an upgrade is offered. Roll back at the tenant level if the eval pass-rate drops.
OpenThree-page runbook for taking your data and weights elsewhere. Yes, we publish it. Yes, the contract obliges us to make it work.
OpenHow it works
Control is built into the platform layer. Every model call passes through a gateway that writes a signed audit event into an append-only chain. Routing is policy-driven, not vendor-driven, so you swap models without rework. Weights are open by default (Llama, Qwen, Mistral). The exit path is contractually required and tested during every onboarding.
Read the full system viewWhat it costs
Control is included in every tier. There's no surcharge — the platform is built this way. The conversation with the discovery agent picks the right tier for your concurrency, residency, and audit posture.