What is Opper?

Opper is one connection to every AI model, with the tools you need to ship AI to real users.

Opper sits between your app and the AI models you use. One API key gets you 700+ models from every major provider, plus the observability and controls to run them in production.

Without it, you end up gluing the pieces together yourself: a separate SDK and key for each provider, and monitoring you add after something breaks. Opper is that whole layer behind one API. Switch models by changing a string, and every call is governed and metered by default.

Build, observe, and control

The Gateway

One API for every model. Send a request, get a response. Switch models with a config change.

Observe & Control

Inspect logs and traces, and manage checks, budgets, model access, retention, and routing.

The Gateway

One API that covers every modality and every major model.

• Any model, same code. Choose openai/gpt-5.5, anthropic/claude-sonnet-4-6, gemini/gemini-3.1-pro-preview, or any of 700+ others by changing a string.
• Beyond text. Send images and PDFs in, and generate images, speech, and video, transcribe audio, and run realtime voice — all through the same API. See Multimodality.
• Use your existing SDK. Already on OpenAI, Anthropic, or Google AI? Point it at Opper and your code stays the same.
• Or write no code at all. Point an app you already use — Claude Code, Cursor, Zed, n8n — at Opper instead. See Apps.
• One key. Replaces a folder of provider credentials.

Observe & Control

Once your AI is in front of real users, you need to understand what it is doing and set clear operating boundaries. Opper lets you inspect calls and manage the rules enforced by the gateway:

	What it does
Data retention	Sets how long request and response content is kept, or keeps none of it.
Spend limits	Caps what an organization can spend in a calendar month.
Model access	Limits which providers, regions, and models can be used.
Checks	Scores responses against your criteria, and blocks or redacts content.
Routing	Chooses which provider a model pool tries first.
Default model	Picks the model for calls that don’t name one, set through the Management API.

The first five live on the organization-level Rules page. Each rule applies to the whole organization or selected projects. You can review rule activity in traces when trace storage is enabled.

Explore Observe & Control →

Where Opper runs

Opper runs in Stockholm, on AWS with a second production environment at evroc, and your data stays in-region. On the Gateway plan it stores only metadata, not your prompts or responses; on Control Plane it stores traces only once you turn tracing on, for the period you choose. Only the model call itself reaches the provider, and you can pin which providers and countries are allowed. The full picture is on the Security page.

Get going

Run your first call, then read the Platform section in order to understand how it all fits.

Quickstart

Five minutes to your first call.

Core concepts

Gateway, projects, Rules, and what happens on a call.

AI Gateway

One API, 700+ models, routing, and EU enforcement.