Anthropic-compatible API

Documentation

Lumosel speaks the Anthropic Messages API. Point any Claude-compatible client at one base URL, use your key, and you're live — Claude Code, cURL, the official SDKs, or any tool that reads the Anthropic environment variables.

Introduction

Everything routes through a single base URL:

Do not append /v1/messages to the base URL — every client adds the path itself. The base URL is just the host.

Authentication

Create a key on the API keys page. It starts with lumo_live_. Send it one of two equivalent ways:

x-api-key: YOUR_API_KEY — the standard Anthropic header (SDKs use this).
Authorization: Bearer YOUR_API_KEY— used by Claude Code's ANTHROPIC_AUTH_TOKEN.

Treat your key like a password — anyone with it can spend your balance. Rotate it from the dashboard if it leaks.

Quickstart

Pick how you want to connect:

~/.claude/settings.json

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://api.lumosel.vip",
    "ANTHROPIC_AUTH_TOKEN": "YOUR_API_KEY",
    "ANTHROPIC_MODEL": "claude-opus-4-8",
    "ANTHROPIC_SMALL_FAST_MODEL": "claude-haiku-4.5"
  }
}

Pin both models so Claude Code never auto-selects one the gateway doesn't serve. Then just run `claude`.

Replace YOUR_API_KEY with your key and claude-opus-4-8 with any model from the table below.

Models

Live list of models the gateway currently serves. Prices are in USD per 1M tokens; you're billed only for what you use, per request.

Model ID	Tier	Input	Output
`claude-opus-4-8`	Most capable	$15	$75
`claude-opus-4.5`	Most capable	$15	$75
`claude-sonnet-4.5`	Balanced	$3	$15
`claude-haiku-4.5`	Fastest	$1	$5
`gpt-5.5`	Most capable	$10	$30

Use the exact Model ID as the model field in your request.

Endpoints

All endpoints are relative to the base URL and mirror the Anthropic API.

POST/v1/messagesCreate a message. Set "stream": true for server-sent events.

POST/v1/messages/count_tokensCount tokens for a request without running it. Not billed.

GET/v1/modelsList the models available to your key.

Errors

The gateway returns standard HTTP status codes with an Anthropic-shaped error body.

`401`	Missing, malformed, or invalid API key.
`404`	Unknown model — check the model ID against the table above.
`429`	Usage window exhausted and balance is empty. Top up or wait for the window to reset.
`503`	Gateway is offline, in maintenance, or the model is temporarily disabled.
`502`	Upstream provider is unreachable. Retries usually recover automatically.

Ready to build?

Grab a key and make your first request in under a minute.

Create an API key