One API key. OpenAI, Anthropic, and Gemini.

Quickstart

Three steps to a first completion. Same bearer token for model discovery and inference.

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.LUCENTA_API_KEY!,
  baseURL: "https://api.lucenta.ai/v1",
});

const res = await client.chat.completions.create({
  model: "claude-sonnet-4-6",
  messages: [{ role: "user", content: "Hello from Lucenta" }],
});

console.log(res.choices[0]?.message?.content);

Local docker: use http://localhost:8000/v1 instead of production. Need credits first? Top up under Billing.

Authentication

Create keys in the dashboard under API Keys. Send the secret on every /v1/* request. Prefer Authorization: Bearer YOUR_API_KEY. Anthropic-style clients (New API, the Anthropic SDK) may send x-api-key instead. If both headers are present they must be the same key. Usage draws from your prepaid balance — if the reserved charge exceeds the balance, Lucenta returns 402 with insufficient_credits.

List models

GET /v1/models returns the public catalog (Anthropic, OpenAI, Gemini) and Lucenta pricing under lucenta. Browse the same list on /models or in the dashboard.

curl https://api.lucenta.ai/v1/models \
  -H "Authorization: Bearer YOUR_API_KEY"

Chat Completions

OpenAI-compatible chat. Required: model and messages. Pass any public model id from GET /v1/models (for example claude-sonnet-4-6).

curl https://api.lucenta.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "messages": [
      {"role": "user", "content": "Hello from Lucenta"}
    ]
  }'

Anthropic Messages

POST /v1/messages takes an Anthropic Messages body and returns content blocks. Use it with the Anthropic SDK by setting its base URL to https://api.lucenta.ai/v1. Auth is Authorization: Bearer or x-api-key. anthropic-version is optional and forwarded when present.

curl https://api.lucenta.ai/v1/messages \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "max_tokens": 256,
    "system": "Be brief.",
    "messages": [
      {"role": "user", "content": [{"type": "text", "text": "Hello from Lucenta"}]}
    ]
  }'

Use a Lucenta public model id from GET /v1/models. Images, PDFs, tools, thinking, and prompt cache are accepted on this endpoint. JSON and stream responses return those content types when the model produces them, including server tools such as web search.

Gemini

POST /v1/beta/models/{model}:generateContent takes a Gemini body —contents with parts — and returns candidates. The model goes in the path, and it is the same public id as everywhere else.

curl https://api.lucenta.ai/v1/beta/models/claude-sonnet-4-6:generateContent \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {"role": "user", "parts": [{"text": "Hello from Lucenta"}]}
    ],
    "generationConfig": {"maxOutputTokens": 256}
  }'

systemInstruction, temperature, topP and maxOutputTokens are honored. Text parts only.

Streaming

Set stream: true on /v1/chat/completions for text/event-stream chunks in the OpenAI SSE format. The same flag on /v1/messages streams Anthropic SSE (event: message_start, content_block_delta, thinking and signature events when requested). Gemini generateContent returns a single JSON response.

curl https://api.lucenta.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-4-6",
    "stream": true,
    "messages": [
      {"role": "user", "content": "Stream a short hello"}
    ]
  }'

Errors

Errors use an OpenAI-style envelope with a Lucenta type. There is no per-key request or token cap — prepaid balance is the only stop. Insufficient credit returns 402.

{
  "error": {
    "message": "Insufficient credits",
    "type": "insufficient_credits",
    "balance": 0.0
  }
}