CHECKING GATEWAY…

One API.
Every AI Model.

Access supported AI models through a unified gateway with flexible routing, streaming, usage visibility, and developer-friendly integrations.

DESIGNED FOR THE REALITIES OF PRODUCTION AI

One key

For supported API surfaces

Model choice

Without integration sprawl

Usage logs

For every metered request

Failover

Across configured upstreams

Model access

Change the model.
Not your code.

One endpoint, one familiar request shape — browse the live gateway catalogue after signing in, then copy the exact id to use.

Illustrative routing view — the live catalogue is in the dashboard.

Request architecture

One control plane.
A wider model surface.

Your application talks to one stable endpoint. The gateway authenticates, routes, meters, and fails over behind it — provider complexity never leaks into your code.

  • OpenAI compatible
  • Anthropic Messages API
  • Streaming support

Production primitives

Infrastructure that stays out of your way.

Practical controls for applications that need to keep moving as models and providers change.

Unified API

OpenAI-compatible Chat Completions and Responses, plus the Anthropic Messages API.

One base URL · one auth scheme

Provider routing

Priority-tier routing with weighted failover, so one upstream incident doesn't stop your app.

Automatic failover · 60s cooldown

Streaming

Token-by-token output streamed over the same interfaces you already use.

SSE passthrough · usage accounting

Usage visibility

Every metered request logged — tokens, model, provider, status, and duration.

Per-request logs · analytics

API key management

Create, scope, and revoke keys from the dashboard without touching provider credentials.

Create · copy once · revoke

Rate protection

Per-key and plan-aware limits enforced before a request leaves the gateway.

120 req/min per key · plan windows

CATALOGUE
7
Seed models in the gateway catalogue
SDK SHAPES
3
cURL, Python, and JavaScript snippets
API SURFACES
4
Chat · Responses · Messages · Images
BASE URL
1
One endpoint for every model

Why a gateway

Less integration.
More optionality.

WITHOUT A GATEWAY
Your application├─ provider SDKs├─ credential handling├─ retry behavior├─ usage aggregation└─ model migration work
WITH LLMFLUX
Your application└─ LLMFlux API   ├─ model catalogue   ├─ upstream routing   ├─ request logging   └─ API key controls

Everything you need. Nothing you don't.

Unified API

OpenAI-compatible Chat Completions and Responses, plus the Anthropic Messages API. One base URL · one auth scheme

Provider routing

Priority-tier routing with weighted failover, so one upstream incident doesn't stop your app. Automatic failover · 60s cooldown

Streaming

Token-by-token output streamed over the same interfaces you already use. SSE passthrough · usage accounting

Usage visibility

Every metered request logged — tokens, model, provider, status, and duration. Per-request logs · analytics

API key management

Create, scope, and revoke keys from the dashboard without touching provider credentials. Create · copy once · revoke

Rate protection

Per-key and plan-aware limits enforced before a request leaves the gateway. 120 req/min per key · plan windows

Getting started

Three steps to your first request.

If you already use the OpenAI or Anthropic SDK, the migration is a base URL and a key — the request shape stays the same.

  1. Create an account

    Sign up and receive starter credit to evaluate the gateway.

  2. Create an API key

    Generate a key from the dashboard — it appears once, then stays masked.

  3. Send a request

    Point your OpenAI or Anthropic client at the gateway base URL.

Flux CLI — Terminal AI Companionv1.2.0

Instant multi-agent mesh, effort reasoning & pipes in your terminal.

Full Guide & Windows Command
$curl -fsSL https://llmflux.dev/install.sh | bash
Install CLI →

Pricing

Start at the endpoint.

Choose the path that fits your current stage, then grow with the infrastructure.

FAQ

Answers for
builders.

For implementation details, see the API reference.

Open API docs

Build once. Ship with every model.

Give your application a durable path into the AI ecosystem without coupling it to every provider.