Unified AI API gateway for SaaS teams

MF Gateway routes AI traffic before it reaches your app.

One OpenAI-compatible edge for provider routing, failover, model discovery, usage telemetry, and cost control across product teams.

Ship through gateway See routing model

live route trace healthy

POST /v1/chat/completions
model: "fast-coder"

gateway.resolve()
  alias       fast-coder
  primary     openai-codex:gpt-5.4
  fallback    anthropic:claude-3-5-sonnet
  timeout     18s

emit telemetry
  provider    openai-codex
  latency     842ms
  status      completed

1 API

OpenAI-compatible entry point for app teams

3 layers

routing, provider health, and usage telemetry

0 lock-in

map public model ids to upstream providers

Routing surface

A gateway layer built for model operations, not demo prompts.

Provider routing

Route every AI request through one operational control plane.

Map model aliases to OpenAI, Anthropic, Google, OpenRouter, or internal adapters without rewriting product code.

Failover

Keep user flows alive when an upstream model stalls.

Centralize retry policy, timeout boundaries, provider health, and fallback choices outside the application layer.

Model discovery

Expose a clean catalog instead of leaking provider-specific inventory.

Publish only the models your teams should use, with context windows, capabilities, and cost posture attached.

Cost visibility

Turn token usage into accountable product telemetry.

Track model, provider, latency, and usage data where engineering teams already inspect service behavior.

Developer workflow

Keep product code boring. Move AI operations into the gateway.

Point SDKs or backend services at the MF Gateway base URL.

Inspect health, usage, and routing behavior from one surface.

API shape

OpenAI-compatible at the edge.

curl https://gateway.example.com/v1/chat/completions \
  --header "Authorization: Bearer $MF_GATEWAY_KEY" \
  --header "Content-Type: application/json" \
  --data '{
    "model": "fast-coder",
    "messages": [
      { "role": "user", "content": "Design a provider failover policy" }
    ]
  }'

Cloudflare deployment

Static Astro output, deployed with Wrangler.

Wrangler docs

Build output lands in dist. The included wrangler.jsonc points Cloudflare assets at that directory for CLI deployment.

FAQ

Operational questions, answered directly.

Is MF Gateway a replacement for provider SDKs?

It is a gateway layer in front of provider APIs. Teams can keep OpenAI-compatible client code while centralizing routing, credentials, and observability.

Does the landing page require backend services?

No. This site is a static Astro build designed for Cloudflare deployment. Product APIs remain in the existing services workspace.

Can teams expose only approved models?

Yes. The gateway model catalog can present stable public ids while upstream provider and model ids stay behind the routing layer.