The Perplexity app for macOS now supports the Model Context Protocol (MCP), unlocking a new level of utility for users. With MCP integration, you can use Perplexity to interact with local apps and services directly from your desktop. That means querying your Apple Notes, setting reminders, or pulling files from Google Drive, all from a single prompt bar.
The core update: MCP gives Perplexity access to external data sources through a standard protocol, similar to how HTTP powers the web. While Perplexity already lets you search with a shortcut, MCP now allows it to do things. Pull files, check calendars, send emails. All by connecting to your system’s data in a secure, structured way.
What MCP Means for Perplexity Users
MCP, short for Model Context Protocol, is an open standard originally proposed by Anthropic. Its goal is simple: make AI tools context-aware by giving them secure access to real data, not just model training. Think of it as the bridge between your AI assistant and the digital tools you use every day.
To get MCP working on Perplexity’s macOS app, you’ll need to install a helper app called PerplexityXPC. This bypasses macOS sandboxing rules that would otherwise block local integrations. Once installed, you can link MCP connectors through the settings panel.
For example, by using a prebuilt MCP server like applescript-mcp, users can run commands such as “Check my Mac calendar” or “Send a note to myself” directly through the Perplexity interface. The integration uses local MCP servers to translate these commands into native system actions.
According to Perplexity’s official documentation, users must first install dependencies like Node.js and retrieve the proper command from the MCP server’s README. Once configured, the server should appear as “Running” under the app’s Connectors section. You’ll then be able to toggle it as a source on the homepage and test it with real commands.
Why the Industry Is Betting on MCP
Anthropic’s vision for MCP is to create a single, reusable way to connect AI models to external systems. In their words, it’s about ending the fragmentation caused by having to write custom integrations for every data source. Instead of relying on one-off solutions, MCP provides an interface that models like Claude 3.5 Sonnet can tap into directly.
Anthropic writes that the Claude Desktop app supports local MCP servers and includes SDKs and prebuilt connectors for popular platforms like GitHub, Slack, Google Drive, and Postgres. Developers can contribute to open repositories or build their own connectors using Claude’s quickstart tools.
The list of early adopters is growing. Companies like Block, Apollo, Zapier, and Salesforce have either implemented MCP or are working with it to simplify data connectivity. Toolmakers like Zed, Replit, and Sourcegraph are also adopting MCP to give their AI agents deeper, real-time access to project context.
Speaking to the protocol’s importance, Dhanji R. Prasanna, CTO at Block, told Anthropic: “Open technologies like the Model Context Protocol are the bridges that connect AI to real-world applications… We are excited to partner on a protocol and use it to build agentic systems.”
Handle with Care
While MCP can streamline and supercharge AI workflows, users need to approach it responsibly. Perplexity’s help pages warn that some integrations rely on open-source tools, which vary in quality and security. Before using one, read the documentation, check for community feedback, and confirm whether the project is actively maintained.
Perplexity notes that users should install MCP connectors carefully and verify the commands involved, especially when they involve local access or require elevated permissions. The company also shared video tutorials and a help center guide to walk users through the setup process.
In short, MCP brings real utility to AI tools like Perplexity. It bridges the gap between static prompts and real-world actions. With growing support across platforms and enterprise systems, it’s not just a feature. It’s the beginning of a broader shift toward context-aware, integrated AI.
