The open source AI gateway.

Give your whole company every model.

One API and one login, scoped by team or key. We handle key management, new models land on day zero, and your own fine-tuned and self-hosted models sit behind the same key.

We support your model OpenAI Anthropic Gemini Meta Qwen GLM Kimi DeepSeek Mistral Your APIs Local Bedrock Azure AI Foundry Fireworks Modal Google

Optimize for quality, speed, and cost.

Quality: the provider serving the model at full precision wins. Same model, every route. The dial changes who answers first.

A degraded provider never becomes an outage.

If the picked provider throttles, the request fails over down the waterfall and the same model answers. Capacity errors only — rate limits and outages, never a silent swap onto something else.

PUT /api/models/{id}/waterfall · org override

One place for every model credential.

Provider keys live at the gateway. People and agents get scoped gateway keys, managed like any other resource.

Make your AI better, faster, cheaper.

The gateway routes. The intelligence layer improves what flows through it.

Enterprise posture

Secure by policy.

Where a request may go is gateway policy, set once for the whole team.

Point your agent here.

Paste the coding-agent prompt into Claude Code, Cursor, or Codex and it wires the endpoint. Prefer to wire it yourself? The curl, JavaScript, and Python tabs are one call away.

# Router Lab hosted gateway
# Dashboard: platform.routerlab.buzz
# API:       https://api.routerlab.buzz/v1
# Paste into Claude Code, Cursor, or Codex.

Ask me for my email. Create a Router Lab account for that
exact address. Store the gateway key as ROUTERLAB_API_KEY.
Never invent an email or a key.

Wire OpenAI-compatible clients to
  base_url = "https://api.routerlab.buzz/v1"
  api_key  = ROUTERLAB_API_KEY

Then GET /v1/models (does not spend). Do not make a paid
completion until I say so. BYOK pass-through is unaffected.

curl https://api.routerlab.buzz/v1/chat/completions \
  -H "Authorization: Bearer $ROUTERLAB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "haiku-class",
    "messages": [{"role":"user","content":"reply with the single word: ok"}]
  }'

import OpenAI from "openai";
const client = new OpenAI({
  baseURL: "https://api.routerlab.buzz/v1",
  apiKey: process.env.ROUTERLAB_API_KEY,
});
const r = await client.chat.completions.create({
  model: "haiku-class",
  messages: [{ role: "user", content: "reply with the single word: ok" }],
});
console.log(r.choices[0].message.content);

from openai import OpenAI
client = OpenAI(
    base_url="https://api.routerlab.buzz/v1",
    api_key=os.environ["ROUTERLAB_API_KEY"],
)
r = client.chat.completions.create(
    model="haiku-class",
    messages=[{"role": "user", "content": "reply with the single word: ok"}],
)
print(r.choices[0].message.content)

OpenAI-compatible. Same shapes for Chat Completions and Responses. Streaming passes straight through. Error bodies name the problem and the recovery.

One endpoint from here.

Point your agents at api.routerlab.buzz/v1 and everything else follows.

Get API key Book a call

The gateway is open source. Run it yourself.