Request a Demo
Documentation

Docs & API reference.

A starting outline for the UltraTrace documentation. Full guides and the complete API reference arrive with general availability — this is a placeholder structure for the first pass.

Overview

UltraTrace is the enterprise control plane for AI observability. It collects usage telemetry across every AI provider you run, enriches it with your business context, and classifies adoption so you know exactly where to expand, elevate, or reallocate. Everything runs inside your own cloud boundary.

This documentation is an early outline. Endpoints, payloads, and parameters shown below are illustrative and subject to change before general availability.

Deploy

UltraTrace ships from the AWS, Azure, and GCP marketplaces as a single-tenant application that runs in infrastructure you control. There are no agents to install and no browser extensions to distribute.

  • Provision from your cloud marketplace into a dedicated project or subscription.
  • Apply the supplied infrastructure template (Terraform modules provided).
  • Confirm the deployment is reachable only within your network boundary.

Connect providers

Connect each AI provider with read-only credentials. UltraTrace pulls sessions, token usage, API volume, and identity data through native admin APIs, audit logs, and exports — then normalizes everything into one schema.

POST /v1/connectors { "provider": "openai_enterprise", "auth": { "type": "admin_api_key", "secret_ref": "kv://ultratrace/openai" }, "mode": "read_only" }

Identity & SSO

Federate with your identity provider over SAML 2.0 or OIDC. Org attributes — department, cost center, location, and more — flow directly from your IdP into business context.

  • Supports Okta, Microsoft Entra ID, and any standards-compliant IdP.
  • Attribute mapping drives the Context Engine automatically.
  • Every context assignment is captured in an effective-dated audit trail.

Context Engine

The Context Engine maps every user to the dimensions your business runs on. Assignments are effective-dated, so historical metrics always reflect the org chart as it was at the time of usage — not as it is today.

PUT /v1/context/assignments { "user_id": "u_9281", "dimensions": { "department": "Engineering", "cost_center": "CC-440" }, "effective_from": "2026-01-01" }

Engagement states

Every user is classified into one of four engagement states, rolled up by any context dimension:

  • Leveraged — deep, sustained, high-value usage.
  • Active — regular usage that hasn't yet reached leveraged depth.
  • Churned — previously active, now lapsed.
  • Never used — provisioned but never engaged.

Diagnostics

Diagnostics translate the engagement picture into three recommended actions: Expand where demand is proven, Elevate where engagement is shallow, and Reallocate where seats sit idle.

API & MCP

Normalized telemetry, context, and adoption intelligence are exposed through an MCP-compatible API so your BI stack, agents, and internal tools can query them directly.

GET /v1/adoption/rollup?dimension=department&state=never_used Authorization: Bearer <token>

Exports & alerts

Export to CSV or your data warehouse, and configure threshold alerts for inactivity, capacity pressure, and usage spikes.

Need something that isn't here yet? Talk to us — we're building the documentation alongside our early-access partners.