Terraform-deployed · Azure native

One gateway for every AI model your company uses

Gateway sits between your internal applications and Microsoft Foundry, Azure OpenAI and third-party providers — giving platform and FinOps teams centralized authentication, per-team cost control and routing governance in one deployable package.

7.4M
requests governed / mo
31%
avg spend reduction
9 days
median time to production
gateway request
POST https://gw.northwind.io/v1/chat/completions
x-gw-app:      internal-copilot
x-gw-tags:     summarization
authorization: Bearer gw_live_3ab8de55…3f52

→ policy: auth:entra-id ✓  rate-limit ✓  tokens ✓
→ route:  Claude Haiku 4 (anthropic/us-east)
→ 200 OK · 640ms · 3,182 tok · $0.0035
→ charged to CC-1002 (Platform Engineering)
no prompt retentionOpenAI-compatible APIper-team chargeback

Three pillars, one control plane

Everything a platform team needs to let the business build with AI without losing track of who is calling what, and what it costs.

Govern

One Entra ID-backed credential per app, scoped model allow-lists, rate limits and token ceilings enforced at the edge before traffic ever reaches a provider.

  • Per-app keys & rotation
  • Token and payload guardrails
  • Full audit trail, no prompt storage

See

Every request is attributed to an app, team and cost center, so FinOps finally gets chargeback-ready numbers instead of a single opaque Azure line item.

  • Spend by team, model, cost center
  • Latency and error SLOs
  • CSV and Power BI export

Control

Route workloads to the cheapest model that fits, fall back automatically when an upstream degrades, and stop runaway spend with hard budget caps.

  • Priority routing rules
  • Automatic provider fallback
  • Soft and hard budget caps
how it deploys

Ships as Terraform. Runs in your own Azure tenant.

Gateway is not another SaaS hop your data has to traverse. You apply a versioned Terraform module into your subscription: API Management, Key Vault, managed identity, Log Analytics and the policy engine, wired in front of Microsoft Foundry and any other model providers you already pay for. Your traffic and telemetry never leave your tenant.

  • terraform apply into an existing subscription — no new vendor egress
  • Entra ID for admin access, managed identity for upstream auth
  • Telemetry lands in your Log Analytics workspace and Power BI
  • Upgrades are module version bumps reviewed in your own PR flow
main.tf
module "ai_gateway" {
  source  = "northwind/azure-ai-gateway/azurerm"
  version = "2.7.1"

  resource_group = "rg-ai-gateway-prod"
  location       = "eastus2"
  apim_sku       = "Premium_v2"

  upstreams = {
    foundry       = { region = "swedencentral" }
    azure_openai  = { region = "eastus2" }
    anthropic     = { via = "private_link" }
  }

  budgets = { org_monthly_usd = 45000, mode = "soft" }
}

Pricing

Platform licensing only — you keep paying your model providers directly at their rates.

Starter

$1,900/ month

For a first platform team standardizing internal AI access.

  • Up to 10 apps
  • 3 upstream providers
  • 30-day telemetry retention
  • Email support
Request a demo

Business

Most popular
$6,500/ month

For multi-team rollouts that need chargeback and policy control.

  • Unlimited apps
  • Unlimited providers & regions
  • 1-year retention + Power BI export
  • Budget caps & fallback routing
  • Shared Slack channel
Request a demo

Enterprise

Custom

For regulated environments with residency and audit requirements.

  • Private Terraform module fork
  • Data residency pinning
  • SSO/SCIM + custom roles
  • 99.95% SLA, named TAM
Talk to sales

See your own AI spend, broken down by team

A 30-minute walkthrough of the console, the Terraform module and what a Foundry-fronted deployment looks like in your tenant.