Quantlix

Quantlix is an AI runtime control plane for production model calls. It is built around a specific problem: teams can approve controls before deployment, but still need to prove what actually happened after an application went live. Quantlix describes its role as sitting in front of every model call and turning it into tamper-evident, time-stamped proof of the approved controls that fired.

Its runtime layer handles policy decisions, redaction, provider attribution, approvals, evals, and stop controls. The product shows a request path in which policy checks, PII redaction, and budget control are enforced before a provider sees the request. It also records audit-grade traces with policy verdicts, provider metadata, and model versions, and it can export audit bundles. The site also describes a hash-chained trace store, per-tenant SHA-256 ledgers, and completed segments that publish Merkle roots to the public Sigstore Rekor log so reviewers can verify anchors without logging in to Quantlix. In addition, it includes budget and reliability visibility, with usage limits, spend signals, and fleet health in the dashboard overview.

Quantlix says the gateway is OpenAI-compatible and supports POST /run and /v1/ routes with deployment binding for chat, messages, and embeddings. It is described as a drop-in gateway for model providers, workflow RAG and agents, and trace export to existing observability tooling. Named providers include Azure OpenAI, OpenAI, Anthropic, Bedrock, Groq, Together, and self-hosted models. The product is also described as offering EU-hosted deployment and self-hosting.

The site names engineers, governance teams, and leaders as audiences for its three surfaces: gateway and traces for engineers, obligation-mapped evidence for governance, and risk, cost, and reliability rollups for leaders.