Documentation

One API for many models

xKiro speaks both the OpenAI Chat Completions and Anthropic Messages formats. Point the SDK you already use at xKiro, change the model ID, and keep the code you have written.

Base URL

https://api.xkiro.com/v1

Auth header

Authorization / x-api-key

Formats

OpenAI · Anthropic

Modalities

Text · Speech · Image

Model ID

vendor/model

Streaming

Server-sent events

Your first request#

curl https://api.xkiro.com/v1/chat/completions \
  -H "Authorization: Bearer $XKIRO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-5.6-sol",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

Start here#

Quick start

Create a key and send your first request in under a minute.

Authentication

API keys, the two accepted headers, and what is public.

API reference

Every endpoint, parameter and response shape.

Capabilities#

Streaming

Server-sent events for both dialects, with cancellation that stops the bill.

Tool calling

Describe your functions and let the model decide when to call them.

Reasoning

Control how deeply a model thinks before it answers.

Vision

Send screenshots, charts and documents alongside your text.

Text to speech

Natural speech, streamed so playback starts early.

Image generation

Create and edit images through an asynchronous job API.

Two dialects, one gateway#

Every model is reachable through both formats. Pick whichever your SDK already speaks — the model does not change, only the shape of the request and response.

OpenAI Chat Completions

POST /v1/chat/completions

Works with the official openai packages and everything built on them.

Anthropic Messages

POST /v1/messages

Works with @anthropic-ai/sdk, anthropic for Python, and Claude Code.

How routing works#

A request names a model, not a provider. xKiro resolves it to a route and, if that route is unavailable, tries the next before returning an error. Two behaviours show up in your logs:

Failover reaches the same model a different way — you get exactly what you asked for. Degradation serves a comparable model when every route for yours is exhausted; you are billed at the price of the model you requested, and the response reports the model you requested. Degradation only ever moves to a model of equal or lower price that supports everything your request needs.

New here? Start with Quick start. Migrating? See Migrate to xKiro.