The Files.com Model Context Protocol (MCP) Server lets your AI application interact with your Files.com site. The AI agent works against your site as if it were an extension of your team, with the same security, auditability, and access controls that apply to any other API client.

Your AI can use the MCP Server to upload and download files, query folders, manage users, and run other site operations.

Files.com offers two ways to run the MCP Server. The Remote MCP Server is cloud-hosted by Files.com and requires no installation. Your AI client connects directly to our hosted endpoint. The Local MCP Server is a Python package you install and run on your own machine or infrastructure. The Remote MCP Server is the right starting point for most customers.

Model Context Protocol

Model Context Protocol (MCP) is an open protocol that standardizes how applications provide context to Artificial Intelligence (AI) Large Language Models (LLMs). MCP provides a standardized way to connect AI models to different data sources and tools.

MCP Servers are lightweight programs that expose specific capabilities through the standardized Model Context Protocol.

Getting Started with the Remote MCP Server

The Remote MCP Server is the fastest way to connect your AI client to Files.com. Files.com hosts and operates the server, so there is nothing to install or maintain. You can connect to multiple Files.com sites at once by adding an entry for each site in your configuration.

To connect, configure your AI client to use the Files.com MCP endpoint at https://<mysite>.files.com/api/ai/mcp/v1/ or https://<mycustomdomain>/api/ai/mcp/v1/.

Your client will need a Files.com API Key to authenticate. Pass the key as the value of the X-FilesAPI-Key HTTP header.

If you are using Claude Desktop, see the Quickstart for Claude Desktop for step-by-step instructions. For other AI clients, refer to your client's documentation for configuring a remote MCP server.

Getting Started with the Local MCP Server

The Local MCP Server is a Python package that you install and run on your own machine or infrastructure. Use this option when your organization requires that no outbound connection be made to a Files.com-hosted server, or when you need to run the server inside a private network.

Install the Local MCP Server from our GitHub repository or by using the pip Python package manager. Then configure your AI client to use it.

Refer to the documentation for your AI application to configure it to use the Files.com MCP Server.

For customers using Claude Desktop, we provide some quickstart instructions to help you get up and running quickly.

Authentication

Both the Remote and Local MCP Servers require a Files.com API Key to authenticate and authorize actions. For details on what actions the MCP Server can perform and how to scope its permissions, see MCP Access Permissions.

Remote vs. Local MCP Server

The Remote MCP Server is the best choice for most customers. Files.com operates the server, so you don't install anything and the server stays up to date on its own.

The Local MCP Server is the right choice when your environment requires on-premises execution or restricts outbound connections to third-party-hosted services.

Both versions expose the same available MCP tools and use the same API Key-based authentication.

Available MCP Tools

Within an MCP server, a "tool" represents a capability or function that the AI can interact with.

Our MCP server has a limited set of available MCP tools today. More tools will be added over time.

Updating the Local MCP Server

The local MCP Server can be updated to the latest version via the pip Python package manager by using the command:

pip install --upgrade files-com-mcp

When using the uv environment manager, use the command:

uv pip install --upgrade files-com-mcp

Removing the Local MCP Server

The local MCP Server can be uninstalled via the pip Python package manager by using the command:

pip uninstall files-com-mcp

When using the uv environment manager, use the command:

uv pip uninstall files-com-mcp