Build on the platform. Keep control of your application.

Use the same client to call OpenAI, Anthropic, and Gemini. Select models in code and inspect their usage and costs together.

Managed credentials remove the need to operate provider accounts for each integration. Bring your own keys when you want providers to bill your account directly.

Upload documents to managed vector stores for chunking and retrieval. Attach those stores to an agent so its file-search tool can retrieve relevant passages.

For applications with their own retrieval pipeline, use the embeddings endpoint to obtain vectors without storing data in Ingram Cloud.

Connect your backend over MCP and let Ingram Cloud run the model-and-tool loop. Configure approvals for actions that need a human decision.

Keep a client-side tool loop when tool execution belongs in your process. The compatibility guide documents both integration paths.

Attribute model usage to your application’s users and customers. Set budgets, inspect run history, and connect usage data to your own reporting.

Your application maps its user IDs to the platform’s scoped runtime instances. Each can have separate context, connections, and spending controls.