Serve your models on OpenRouter

One integration puts your models in front of the developers already building on OpenRouter.

10M+ developers send requests through OpenRouter, and we route them across 80+ providers by performance and price. Apply once, then track review, onboarding and setup in one place.

We're working through a large backlog of applications, and we review providers with proprietary models first.

Already applied? Check your status

How it works

  1. Step 1

    Apply
    About 15 minutes. Tell us about your infrastructure, endpoints, models and data policies.
    You
  2. Step 2

    Review
    We score your API compatibility, endpoint reliability, pricing and performance. We decide on a rolling basis.
    OpenRouter
  3. Step 3

    Intro call
    We sign a mutual NDA, then meet on a call.
    Both
  4. Step 4

    Security and terms
    You answer the security questionnaire, upload your security documents and accept the provider terms online.
    You
  5. Step 5

    Go live
    We set up the integration together and send test traffic to check latency, throughput and error handling. Then you're live, and production requests reach your models, routed by performance and price.
    Both

Before you apply

If you don't meet them yet, we can't review your application.

SOC 2 and zero data retention help, but neither is required.

View the full technical requirements

  • An OpenAI-compatible API

    Your /chat/completions endpoint supports streaming and returns usage tokens for both stream and non-stream requests.

  • A models endpoint

    Your /models endpoint lists each model with its pricing, context length, max output tokens, supported features and datacenter locations.

  • Monthly invoicing to OpenRouter
    Required

    You invoice OpenRouter monthly for inference. We don't onboard providers on other payment terms.

  • A published privacy and data policy

    It says whether you log prompts, how long you keep data, and whether you train on it.

  • A work email on your company domain

    We can't accept applications from personal addresses such as Gmail or Outlook.

What you get on OpenRouter

One API for developers

10M+ developers already use OpenRouter's OpenAI-compatible API. Once you're live, they can call your models through that same API, with no provider-specific integration.

Performance-based routing

By default, we route requests by latency, throughput, uptime and price, so providers that perform well get proportionally more traffic.

Public performance metrics

Every model page shows your TTFT, throughput and uptime next to other providers, so developers see how you compare when they choose.

Automated payments

You send OpenRouter one invoice a month for usage. We reconcile token counts automatically.

Uptime monitoring

We monitor every endpoint's reliability. Stay at 95% uptime or higher to keep standard routing priority.

Geographic routing

List your datacenter locations in your /models endpoint. We use them to honor developers' region and data residency requirements.

FAQ

Can't find what you need?
Reach out to [email protected]

Submit your application

We give priority to providers that fill gaps in our network. We can't accept every provider.