Unified AI API documentation

Unified AI API documentation, without hidden boundaries.

Use familiar SDKs for common requests. Choose a native endpoint when a provider capability cannot be translated without losing meaning.

Quickstart

Point a supported OpenAI client at the EveryModelAPI base URL and use a project-scoped key.

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.EVERYMODEL_API_KEY,
  baseURL: "https://api.everymodelapi.com/v1",
});

const response = await client.responses.create({
  model: "openai/gpt",
  input: "Summarize the release notes.",
});

Authentication

Send the API key in the Authorization: Bearer header. Keep production, staging, batch jobs, and individual developer access on separate keys.

Model IDs

Model IDs use a provider-qualified pattern such as openai/gpt. Stable aliases will have published update policies; version-pinned identifiers will remain available where upstream terms permit.

Errors and request IDs

Every response will include a platform request ID. Preserve it in logs and support tickets. Errors will distinguish authentication, account policy, rate limit, routing availability, upstream rejection, and gateway failure.

The API is in developer preview. Endpoint details may change before keys are issued, and all examples currently describe the intended public contract.