Claude, Anthropic's AI assistant, is now available to run directly through SAP's enterprise AI infrastructure, giving large organizations a way to use Claude while keeping all data and requests within their own compliance frameworks. This integration, demonstrated by SAP engineers, uses an open-source translation layer called LiteLLM to convert Claude API requests into the format SAP's systems expect, then routes them back through the company's governance controls before reaching the underlying model.

Why Are Enterprises Adding This Extra Layer Between Claude and Their Data?

For many large organizations, using AI directly from Anthropic's public API creates a governance problem. Data leaves the company's network, audit trails become fragmented, and compliance teams lose visibility into what's happening. The SAP integration solves this by inserting a proxy server that sits between Claude and the user, capturing every request and response within the company's own infrastructure.

This matters because enterprises operate under strict regulatory requirements. Financial services firms must comply with data residency rules. Healthcare organizations need to ensure patient data never leaves their systems. Government contractors face classified data restrictions. By routing Claude through SAP's Generative AI Hub, these organizations can enforce their own access controls, encryption policies, and audit logging before any request ever reaches Anthropic's servers.

How Does the Technical Setup Actually Work?

The integration uses three layers. Claude Code or any AI agent sends a standard Anthropic API request. LiteLLM, running as a proxy server on the company's infrastructure, receives that request and translates it into SAP AI Core's format, adding the company's authentication credentials. SAP AI Core then enforces the organization's compliance policies and routes the request to the underlying Claude model, whether that's Claude Sonnet 4.5 or Claude Opus 4.7.

The setup requires a few prerequisites. Organizations need an SAP Business Technology Platform account with AI Core provisioned, models already deployed in their resource group, and a service key for authentication. LiteLLM can run locally on a developer's machine for testing, or containerized with Docker for a persistent, production-ready setup that includes a Postgres database for storing keys and usage logs.

Steps to Deploy Claude Through SAP's Enterprise AI Hub

• Set Up Infrastructure: Provision an SAP BTP account with an AI Core service instance on the extended plan, download your service key, and deploy the models you want to use in your resource group.
• Configure LiteLLM: Pull the LiteLLM Docker image, download the official Docker Compose file, and create a config.yaml file that maps model names like "claude-opus-4-7" to SAP's internal model identifiers like "sap/anthropic--claude-4.7-opus."
• Set Environment Variables: Create a.env file with your LiteLLM master key, salt key, SAP service key credentials, and resource group identifier so the proxy can authenticate with SAP.
• Deploy and Test: Run docker compose up to start the proxy, wait 30 to 40 seconds for the containers to initialize, then send a test API request to confirm the full chain is working end-to-end.
• Connect Your AI Agent: Point Claude Code or any other AI agent to your LiteLLM proxy by setting the ANTHROPIC_BASE_URL environment variable to your proxy's address, then specify the model you configured.

Once deployed, all Claude traffic flows through your SAP tenant's access controls, compliance policies, and audit systems. The integration is transparent to the user; Claude Code responds normally, but every interaction is now governed by your organization's governance framework.

What Does This Mean for the Broader AI Market?

This integration reflects a larger trend in enterprise AI adoption. As organizations move beyond pilots and experiments, they're demanding more control over where their data goes and how it's processed. Public cloud AI APIs are convenient, but they don't fit the compliance and governance requirements of regulated industries. By making Claude available through enterprise infrastructure platforms like SAP, Anthropic is signaling that it understands this constraint and is willing to work within it.

The use of LiteLLM as a translation layer is particularly significant. LiteLLM is an open-source project that speaks multiple AI API formats, including OpenAI's, Anthropic's, and others. This means the same infrastructure pattern could theoretically support multiple AI providers, giving enterprises the flexibility to mix and match models while maintaining a single governance layer. SAP engineers explicitly noted that any AI coding agent compatible with LiteLLM could integrate with SAP AI Core using this same approach.

"Using LiteLLM as a universal proxy layer you can route for example Claude Code through SAP Generative AI Hub," explained the SAP engineer who documented the integration.

SAP Community Technology Blog Post

For organizations already invested in SAP's ecosystem, this integration removes a major friction point. Teams that have standardized on SAP for enterprise resource planning, customer relationship management, or supply chain management can now add Claude-powered AI capabilities without introducing a new vendor or breaking their existing governance structure. The integration also works with Claude Sonnet 4.5 and Claude Opus 4.7, giving enterprises a choice between a faster, more cost-effective model and a more capable reasoning model.

The broader implication is that enterprise AI adoption is moving away from the "use the public API" model and toward "embed AI within our existing infrastructure" model. This shift favors AI providers that are willing to integrate deeply with enterprise platforms and respect organizational governance requirements. For Anthropic, making Claude available through SAP is a strategic move to compete for enterprise adoption in regulated industries where governance and compliance are non-negotiable.