unifai — one API for all your AI

Providers

Configure each LLM provider once. Keys are encrypted at rest. Enable a provider to make it callable through the gateway.

Client keys

Keys your other apps use to authenticate to unifai. Send as Authorization: Bearer <key>.

New key name (e.g. "billing-app", "internal-tools")

Name	Key	Last used	Created

Account

Change your dashboard password.

Current password

New password (min 8 characters)

Price discovery admin

When a call uses a model with no saved price, unifai looks it up on the internet (fallback: this default LLM) and saves it. Also used by the Pricing page's "Refresh & verify".

Default LLM — provider

Default LLM — model

Pick a capable, cheap model (it just answers price lookups). Uses the admin account's provider key. Leave blank to disable.

Gateway admin

How long the gateway waits for a provider to respond before timing out a call. Streaming calls use this only to connect; then they run as long as data keeps flowing.

Upstream timeout (seconds)

Range 1–600s. Raise it for slow reasoning models on non-streaming calls (default 60s).

Playground

Mirrors an external app: sends your unifai client key to the gateway (POST /v1/chat). unifai then uses the provider key you saved under Settings — you never enter a provider key here.

unifai client key

The key an external app sends as Authorization: Bearer …. Create one under Settings → Client keys.

Provider

Model

System prompt (optional)

Prompt

Calling app (for reports)

Functionality (for reports)

Temperature (optional)

Max tokens (optional)

Response

Response will appear here.

Equivalent request — copy this into your app

—

Usage reports

Actual usage across providers, models and calling apps.

Range

From

To

Provider

App (calling app)

Client key

Model

Breakdown

Key	Reqs	Tokens	Cost	Avg ms

Daily spend (estimated cost per day)

Show recent requests

Time	Provider	Model	App	Functionality	Tokens	Cost	ms	Status

Usage limits

Cap spend by budget over a period. A hard limit makes the gateway return a "limit reached" error once the budget is hit; a warning limit emails you but keeps serving requests.

Applies to

Period

Budget (USD)

Type

Applies to	Period	Budget	Spent	Type	Enabled

Model prices

All saved prices (USD per 1M tokens). Unknown models get looked up automatically via your default LLM; "Refresh & verify" re-checks every saved price.

Saved prices

Provider	Model	Input $/1M	Output $/1M	Source	Updated

Set / override a price manually

Use this for models the default LLM can't find (e.g. brand-new ones). Prices are USD per 1M tokens.

Provider

Model

Input $/1M

Output $/1M

Users

Invite people to unifai. Each user sets up their own provider keys and sees only their own usage.

Invite a user

Create an invite link and send it to them. They click it, set a password, and configure their own keys.

Email (optional — locks the invite to this email)

Role

People

Email	Name	Role	Status	Joined

Pending invites

Email	Role	Created	Expires	Link

Using the API

Point any app at unifai. One endpoint, any provider.