Keep your best model for the hard parts.

1. 1

   Create a key

   In the app. It is shown once, so copy it when it appears.

2. 2

   Point your client at it

   Give it the base URL and the key, as in the samples below.

3. 3

   Add your own models

   Optional. Add a provider key and put models on your ladder; without them the house model answers.

export ANTHROPIC_BASE_URL="https://www.batonapi.com"
export ANTHROPIC_AUTH_TOKEN="YOUR_BATON_KEY"
export ANTHROPIC_API_KEY=""
claude

# pip install openai
from openai import OpenAI

client = OpenAI(
    base_url="https://www.batonapi.com/v1",
    api_key="YOUR_BATON_KEY",
)

completion = client.chat.completions.create(
    model="baton",
    messages=[{"role": "user", "content": "Explain what a mutex is in one sentence."}],
)

print(completion.choices[0].message.content)

// npm install openai
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://www.batonapi.com/v1",
  apiKey: "YOUR_BATON_KEY",
});

const completion = await client.chat.completions.create({
  model: "baton",
  messages: [{ role: "user", content: "Explain what a mutex is in one sentence." }],
});

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

# pip install anthropic
import anthropic

client = anthropic.Anthropic(
    base_url="https://www.batonapi.com",
    api_key="YOUR_BATON_KEY",
)

message = client.messages.create(
    model="baton",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain what a mutex is in one sentence."}],
)

for block in message.content:
    if block.type == "text":
        print(block.text)

curl https://www.batonapi.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_BATON_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "baton",
    "messages": [{"role": "user", "content": "Explain what a mutex is in one sentence."}]
  }'

Read the setup guide

Which models does it work with?

Anything you can reach with an Anthropic, OpenAI, Google or OpenRouter key, plus any OpenAI-compatible endpoint you run yourself. You put the models in order and Baton works down the list.

How does it decide how hard a request is?

By rules, not by another model reading your prompt: the length of what you asked, words that signal hard or easy work, whether tools are attached, whether you asked for reasoning, and how much context came along. You can override the grade per request with a header.

What counts against a budget?

Each model on your ladder can have a daily budget in dollars. Spend is counted from the token usage the provider reports and the prices set on that model, and resets at midnight UTC. A rate-limit error also moves work down until the provider's retry time has passed.

Does a conversation stay on one model?

Yes, by default. Prompt caches and thinking blocks belong to one model, so a conversation stays where it is until its model runs out of budget or is rate limited. You can turn this off in settings.

What happens to my provider keys?

They are encrypted before they are stored and used only to call that provider for you. The request log keeps the model, token counts, cost and timing of each request, never the prompt or the reply.

What does it cost?

Requests that go to your own provider keys are billed by the provider, as before. The house model is paid from prepaid credit.

Is a router always the right call?

No. If one model at a lower reasoning effort gets you through the day, that is simpler and keeps a single prompt cache. Baton is for when it still runs out.