Core Concepts

Before you integrate AnyInt, it helps to think about the platform in terms of a few core ideas.

1. One gateway, multiple model families

AnyInt is one platform, but it exposes multiple route families:

• OpenAI-compatible
• Anthropic-compatible
• Gemini-compatible
• media generation
• AI Music

That means "one platform" does not mean "one request shape". It means one operational layer across multiple AI surfaces.

2. Easier access is part of the product

AnyInt is not only a relay. One of its core purposes is reducing the effort required to use advanced AI models.

That includes:

• fewer provider-specific integrations
• fewer billing surfaces
• one place to manage keys and access
• one docs set for multiple workflows

3. Model choice is a product problem

Many users know what task they want to solve, but they do not know which model is the best fit. This is why model selection should not be treated as a purely manual decision.

In many real cases:

• multiple models can achieve the same acceptable result
• the highest-priced model is not always necessary
• the biggest challenge is matching task requirements to the right model and price point

4. Smart Routing (Coming Soon) is about better defaults

Smart Routing (Coming Soon) should be understood as a model-selection layer, not just a transport feature.

Its value is:

• choosing models more appropriately
• avoiding unnecessary premium-model usage
• improving the balance between quality, latency, and cost

5. Model IDs are the source of truth

Do not guess model IDs. Fetch them from GET /openai/v1/models or use the list available to your account. Availability can differ by account, billing mode, plan, or purchase configuration.

6. API keys connect access and billing

API keys are the main authentication mechanism for platform access. Depending on the route family you call, the header format may differ:

• Authorization: Bearer <key>
• x-api-key: <key>

Keys also connect to billing mode, because requests may consume prepaid balance, subscription entitlement, or team-managed access policy.

7. Async tasks are normal for media workflows

Not every route returns a final artifact immediately. Media and AI Music workflows commonly follow this pattern:

1. create a task
2. receive a task ID
3. poll a query endpoint or wait for a webhook
4. download the generated artifact

The mental model to keep

Think of AnyInt as:

• an easier way to access advanced AI models
• a platform for better model selection
• a gateway for multiple API families
• a control surface for billing, keys, and policy
• a workflow layer for async media and music generation