Overview

Ada AI is a model gateway and control plane. Bring your provider keys, route traffic through one endpoint, and keep explicit control over model selection, priority, and failover.

Ada AI gives you one OpenAI- and Anthropic-compatible API across the model providers you already use. You bring the provider keys; the proxy routes your traffic across them in priority order with automatic 5xx failover.

Dashboard: ada.ai
API base URL: https://api.ada.ai/v1

What it does

One endpoint, many providers. A single sk-rc-… proxy key addresses every model you've enabled, no matter which provider hosts it. The request shape is identical to OpenAI's, so existing SDKs and tools work without changes — point base_url at Ada AI and use the proxy key as the API key.
Explicit routing. The proxy picks an upstream by matching the requested model against each upstream's synced model list. When several upstreams expose the same model, the one with the higher priority number is tried first. It never silently substitutes a different model.
Automatic failover. If an upstream returns a 5xx, the proxy retries the next matching upstream automatically.
Team access. Organizations let multiple people share one pool of upstreams. Each proxy key is scoped to exactly one owner — you, or one organization — fixed when the key is minted.
Usage signals. Every request records tokens, latency, time-to-first-token, provider, and status, available in the dashboard and via GET /me/usage.

How a request flows

your app  ──sk-rc-…──►  Ada AI (api.ada.ai/v1)
                            │  matches model → enabled upstreams
                            ├──►  upstream #1 (priority 10)  ──5xx──┐
                            ├──►  upstream #2 (priority 1)  ◄──────┘  failover
                            └──►  response (OpenAI or Anthropic shape)

Where to start

Rocket

Quickstart

Terminal

Coding agents

Point Pi and OpenCode at Ada AI so every request flows through your proxy key and routing rules.

MessageSquare

Chat completions

The OpenAI-compatible /v1/chat/completions endpoint with streaming and SDK examples.

MessagesSquare

Anthropic messages

Use /v1/messages to talk to any model with the Anthropic SDK.

Route

Models & routing

How model discovery, priority ordering, and failover work.

Key

API keys

Mint, scope, rotate, and revoke proxy keys.

Users

Organizations

Share upstreams with a team under one account.

Not sure which API to use?

If you...	Use
Already call OpenAI's `/v1/chat/completions`	`/v1/chat/completions` — drop-in replacement
Already use the Anthropic SDK's `messages.create`	`/v1/messages` — same shape, any provider's model
Want to list what's available	`GET /v1/models` — see Models & routing
Are building a CLI tool	CLI authentication — the browser-assisted device flow
Want to drive a coding agent	Coding agents — Pi and OpenCode, configured in minutes