Skip to content
baton
Open app

/changelog

Baton changelog

What changed in Baton, newest first.

  1. Rate limits

    The API now limits how fast one workspace can send requests, and sign-in limits how many workspaces one address can create.

    • 120 requests a minute per workspace. Over that, the API answers 429 with a Retry-After header.
    • 5 new workspaces an hour per address when signing in without a wallet.

    Errors and what to do about them

  2. Savings report

    A new page in the dashboard shows what the ladder saved: every request priced as if your top model had answered it, next to what you paid.

    • Over the last 7 or 30 days, by model and by day.
    • An estimate, worked out from the prices on your models. It needs prices on the top one.

    Open Savings

  3. Baton is live

    One API key for every model. Requests are graded by rules, hard ones stay on your best model, easier ones move down a ladder of cheaper models as daily budgets run down, and the house model answers when the rest are out.

    • OpenAI-compatible and Anthropic-compatible endpoints, streaming included, with translation between the two.
    • Your own keys for Anthropic, OpenAI, Google, OpenRouter, or any OpenAI-compatible endpoint.
    • A dashboard for the ladder, keys, the request log and a playground.
    • Every reply names the model that wrote it in the x-baton-model response header.
    • An MCP server for agents that would rather call a tool than an API.

    Quickstart