What is the New Enterprise Capability?
Anthropic is rolling out a significant upgrade for its Claude Enterprise plan that allows the AI to securely connect to and interact with a company's own internal data sources. This capability, often referred to as 'tool use' or function calling, fundamentally
changes how employees can leverage AI. Instead of manually copying and pasting information into a chat window, users can now grant Claude permission to directly query databases, access files in cloud storage, or pull information from SaaS platforms like Salesforce. The initiative is powered by an open-source standard Anthropic developed called the Model Context Protocol (MCP), which is designed to create a universal and secure bridge between AI assistants and business data systems.
From Chatbot to Integrated Co-Worker
This marks a pivotal shift from the AI as a conversational partner to the AI as an active participant in workflows. Think of it as the difference between emailing a new hire a spreadsheet versus giving them a login to the company's live financial system. The new features enable Claude to perform multi-step tasks that require access to real-time, proprietary information. For example, an employee could ask Claude to 'analyze sales performance in the last quarter for our top three products and create a slide deck with the key trends'. To do this, Claude could query a sales database like Snowflake, analyze the results, and then use a design tool to generate the presentation slides, all within a single request.
How It Works: APIs and Connectors
The magic behind this connection lies in APIs and a framework of 'tools' or 'connectors'. Anthropic provides a way for developers to define external functions that Claude can call. When a user makes a request, Claude determines if one of its available tools can help. If so, it requests to use that tool. Crucially, your company's systems execute the function—not Anthropic's. This means your data and credentials remain within your control. Anthropic and its partners are building a catalog of pre-built connectors for popular services like Google Drive, Salesforce, and databases like Postgres, while also allowing companies to build their own custom connectors for bespoke internal systems.
The All-Important Security Question
For any business, giving an external system access to internal data raises immediate security and privacy concerns. Anthropic has built its enterprise offering with this at its core. Data sent through Claude Enterprise is not used for training models by default. The platform employs strong encryption for data both at rest (AES-256) and in transit (TLS 1.2+). Furthermore, Claude Enterprise includes critical governance features like Single Sign-On (SSO) to manage user identity, audit logs to track usage, and fine-grained access controls that allow administrators to decide which users or roles can use specific data connectors. Companies can even opt for zero data retention for some API uses, ensuring sensitive information isn't stored after a request is processed.
The Competitive Race for the Enterprise
Anthropic's move is a direct response to a fiercely competitive enterprise AI market. Its primary rival, OpenAI, offers similar capabilities with its ChatGPT Enterprise plan, which also allows for connecting to company data and has a large ecosystem of integrations. Microsoft's Copilot is also deeply embedded in the enterprise through its integration with Microsoft 365 and Azure. While both Claude and ChatGPT offer the core security and administrative controls enterprises demand, the choice often comes down to workflow fit. Claude has often been favored for tasks involving long, complex documents and high-quality code generation, and this deeper data integration aims to solidify its position as the go-to assistant for knowledge-heavy and technical work.
