$api docs

RedRoute (beta) keys aren't self-serve. The operator hands out a one-time Discord token in our server, and you redeem it for an API key on your account.

1. Join the Discord: https://discord.gg/your-invite
2. Ask the operator for a token in #access. You'll get a string like rr_tkn_xxxxxxxx.
3. Sign in at https://redroute.pro/login with your email.
4. Paste the token at https://redroute.pro/redeem.
5. Copy the API key shown once — that's your rr_live_....

Lost the key? You can revoke and mint a new one from the dashboard after sign-in. Tokens are single-use and permanently bound to the account that redeems them.

curl https://redroute.pro/v1/chat/completions \
  -H "Authorization: Bearer rr_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.5",
    "messages": [
      { "role": "user", "content": "Reply with exactly: redroute-ok" }
    ],
    "temperature": 0,
    "max_tokens": 32
  }'

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: "rr_live_YOUR_KEY",
  baseURL: "https://redroute.pro/v1"
});

const res = await client.chat.completions.create({
  model: "gpt-5.5",
  messages: [{ role: "user", content: "Reply with redroute-ok" }],
  temperature: 0,
  max_tokens: 32
});

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

Set stream: true. For usage accounting, include stream_options.include_usage: true when your client supports it.

curl https://redroute.pro/v1/chat/completions \
  -H "Authorization: Bearer rr_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4-8",
    "stream": true,
    "stream_options": { "include_usage": true },
    "messages": [
      { "role": "user", "content": "Say hello in one sentence." }
    ],
    "max_tokens": 128
  }'

Send images as image_url content parts in the messages array. Both base64 data URLs and remote URLs are supported. Max request body size is 12 MB.

Supported formats: image/png, image/jpeg, image/webp, image/gif.

{
  "model": "gpt-5.5",
  "messages": [{
    "role": "user",
    "content": [
      { "type": "text", "text": "What is in this screenshot?" },
      { "type": "image_url", "image_url": { "url": "data:image/png;base64,..." } }
    ]
  }],
  "max_tokens": 256
}

Anthropic-style image blocks on /v1/messages are automatically converted to image_url before forwarding upstream.

Generate images with gpt-image-2 or gpt-image-1-mini. Billing is per image, not per token.

{
  "model": "gpt-image-2",
  "prompt": "A red panda in a space suit, digital art",
  "size": "1024x1024",
  "n": 1
}

Or try it interactively at /images.

Generate speech with gpt-4o-mini-tts or gemini-2.5-flash-preview-tts. Billing is either flat per request or token-estimated depending on the model.

curl https://redroute.pro/v1/audio/speech \
  -H "Authorization: Bearer rr_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  --output speech.mp3 \
  -d '{
    "model": "gpt-4o-mini-tts",
    "voice": "nova",
    "input": "RedRoute can generate speech through the audio route."
  }'

Or try it interactively at /Audio.

Transcribe audio with whisper-1. Send a multipart form with a file field (up to 25 MB) and optional language, prompt, temperature, and response_format (json, text, verbose_json, srt, vtt). Billing is per audio minute.

curl https://redroute.pro/v1/audio/transcriptions \
  -H "Authorization: Bearer rr_live_YOUR_KEY" \
  -F model="whisper-1" \
  -F file="@meeting.mp3" \
  -F response_format="json"

Connect a WebSocket for live speech-to-speech with gpt-realtime-2025-08-28. Authenticate with Authorization: Bearer rr_live_... or the browser-style openai-insecure-api-key.<key> subprotocol. Text tokens bill at the listed rate; audio tokens at $32/$64 per 1M. Each session reserves $1.00 of credit, refunds the unused part on disconnect, and closes automatically at the budget or after 30 minutes.

import WebSocket from "ws";

const ws = new WebSocket(
  "wss://redroute.pro/v1/realtime?model=gpt-realtime-2025-08-28",
  { headers: { Authorization: "Bearer rr_live_YOUR_KEY" } }
);

ws.on("message", (data) => console.log(JSON.parse(data)));
ws.on("open", () => {
  ws.send(JSON.stringify({
    type: "response.create",
    response: { modalities: ["audio", "text"], instructions: "Say hello." }
  }));
});

Submit PixVerse video renders asynchronously, then poll the RedRoute job id and stream the stored MP4 when completed.

POST /v1/videos
GET /v1/videos/{id}
GET /v1/videos/{id}/content

Or try it interactively at /studio.

Drop this into ~/.config/opencode/opencode.json (or your project's opencode.json) to use RedRoute as a custom provider. The modalities field is the important bit — without it OpenCode silently strips images and tells the model "this model does not support image input."

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "redroute": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "RedRoute",
      "options": {
        "baseURL": "https://redroute.pro/v1",
        "apiKey": "{env:REDROUTE_API_KEY}"
      },
      "models": {
        "gpt-5.5": {
          "name": "GPT-5.5",
          "attachment": true,
          "reasoning": true,
          "modalities": { "input": ["text", "image"], "output": ["text"] },
          "limit": { "context": 400000, "output": 16384 },
          "options": { "reasoningEffort": "medium", "reasoningSummary": "auto" }
        },
        "claude-fable-5": {
          "name": "Claude Fable 5",
          "attachment": true,
          "modalities": { "input": ["text", "image"], "output": ["text"] },
          "limit": { "context": 1000000, "output": 16384 }
        },
        "claude-opus-4-8": {
          "name": "Claude Opus 4.8",
          "attachment": true,
          "modalities": { "input": ["text", "image"], "output": ["text"] },
          "limit": { "context": 200000, "output": 16384 }
        },
        "deepseek-v4-flash": {
          "name": "DeepSeek V4 Flash",
          "attachment": false,
          "reasoning": true,
          "modalities": { "input": ["text"], "output": ["text"] },
          "limit": { "context": 1000000, "output": 384000 },
          "options": { "reasoningEffort": "medium" }
        },
        "deepseek-v4-pro": {
          "name": "DeepSeek V4 Pro",
          "attachment": false,
          "reasoning": true,
          "modalities": { "input": ["text"], "output": ["text"] },
          "limit": { "context": 1000000, "output": 384000 },
          "options": { "reasoningEffort": "medium" }
        },
        "MiniMax-M3": {
          "name": "MiniMax M3",
          "attachment": false,
          "reasoning": true,
          "modalities": { "input": ["text"], "output": ["text"] },
          "limit": { "context": 1000000, "output": 32768 },
          "options": { "reasoningEffort": "medium" }
        },
        "glm-5.1": {
          "name": "GLM-5.1",
          "attachment": false,
          "reasoning": true,
          "modalities": { "input": ["text"], "output": ["text"] },
          "limit": { "context": 198000, "output": 128000 },
          "options": { "reasoningEffort": "medium" }
        },
        "qwen3.7-max": {
          "name": "Qwen3.7 Max",
          "attachment": false,
          "reasoning": true,
          "modalities": { "input": ["text"], "output": ["text"] },
          "limit": { "context": 1000000, "output": 64000 },
          "options": { "reasoningEffort": "medium" }
        }
      }
    }
  }
}

Then export your key and run opencode:

export REDROUTE_API_KEY=rr_live_YOUR_KEY
opencode

Pick a RedRoute model with /models. Drag images into the terminal — with modalities.input including "image", OpenCode forwards them as image_url parts and RedRoute passes them through to the upstream model.