Google, Apple, Samsung, and virtually all other major tech companies promise a future in which artificial intelligence will take over your phone and computer. However, today’s newfangled AI assistants are not as smart as their names suggest. Anthropic, the developer of the Claude chatbot, has a new tool that promises to give any assistant access to any application. If the future is indeed based on “agent-based artificial intelligence,” then this is the model Apple Intelligence or Gemini will use to collect data from any app on your devices.
The tool is called the Model Context Protocol, which Anthropic announced on Monday, making it open and free for any developer to use. This protocol will create an information channel between a data source, even application data, and an AI assistant. It is designed to be used by AI agents such as Anthropic’s Claude desktop app. Anthropic said that this connection is “secure” because instead of disclosing their app’s data, third-party developers can create an “MCP client,” also known as a standalone AI application, to connect to their private servers.
While this could ostensibly be used by companies to make it easier for AI to access internal data, it could also offer a new standard for how apps provide their valuable data to AI assistants. Currently, AI apps such as Google’s Gemini on Android and iOS can access the tech giant’s own Workspace apps if users give it permission. That means it can dig through your Gmail to find lost messages or turn any text booking into an event on Google Calendar. Currently, Gemini doesn’t have access to third-party apps except through extensions.
Alex Albert, head of public relations at Claude, claims that developers usually have to work directly with Google or any other AI assistant developer to create their own connection between the AI and its data. MCP claims to solve this problem with a free protocol that can work between an AI assistant and any application or data source.
Albert showed Claude how to work directly with GitHub to create a new repository using MCP. He claimed that using MCP, it took “less than an hour” to create an AI integration with applications. The company shared some repositories of sample MCP servers that it claims should help developers get started.
Although the protocol is open source, Anthropic still claims that the latest Claude 3.5 Sonnet model is better for creating data channels between servers and AI. Otherwise, according to Anthropic, this tool will replace today’s “fragmented” integrations with something more “robust.” Currently, according to Anthropic, software companies such as Zed, Replit, Codeium, and Sourcegraph have connected to MCP to integrate artificial intelligence.
Apple has promised similar cross-software AI functionality for Apple Intelligence. This feature, along with a rejuvenated Siri, may appear in early spring next year. There may not be room for MCP in the company’s famous secure garden in Cupertino, California, but other, smaller AI agent developers may be looking at it or some other similar protocol.
Either way, the largest AI developers will put pressure on app makers to give AI assistants access to app data. While Apple claims that its private cloud computing will keep all your personal data safe, we’ll have to see if this public data transfer will not become another privacy SNAFU.
