Edit

MCP server for AI agents

DevFlow includes a Model Context Protocol (MCP) server that exposes 50+ structured tools for AI agent integration. This allows AI assistants like Claude, GitHub Copilot, and others to inspect, interact with, and debug running MAUI applications.

Important

DevFlow is experimental and will change between releases.

What is MCP

Model Context Protocol (MCP) is a standard for AI tool integration. It defines a structured way for AI agents to discover and call tools exposed by a server. The DevFlow MCP server exposes MAUI app capabilities as structured tools that AI agents can call, enabling AI-assisted development, testing, and debugging workflows.

Start the MCP server

Use the mcp command to start the MCP server:

maui devflow mcp

This starts a stdio-based MCP server. AI agents connect to the server through stdio transport, discovering available tools and calling them to interact with the running MAUI app.

Available tools

The MCP server exposes tools organized into the following categories:

Category Description
Agent Agent status and connected agent listing
Assert UI assertions for testing
CDP Chrome DevTools Protocol for Blazor WebViews
Interaction Tap, fill, scroll UI elements
Logs Application log retrieval
Navigation Shell and page navigation
Network HTTP request/response monitoring
Platform Platform and device information
Preferences App preferences read/write
Property Element property inspection and mutation
Storage File storage root discovery and sandboxed file management
Query CSS selector-based element queries
Recording Interaction recording
Screenshot App screenshot capture
Sensor Device sensor data
Tree Visual tree inspection

Configure with AI agents

The MCP server uses stdio transport, which is the standard transport mechanism supported by most AI agent frameworks. Add the DevFlow MCP server to your AI tool's configuration:

{
  "mcpServers": {
    "maui-devflow": {
      "command": "maui",
      "args": ["devflow", "mcp"]
    }
  }
}

Refer to your AI tool's documentation for the specific configuration file location and format.

Example workflow

An AI agent connected to the DevFlow MCP server can perform end-to-end automation tasks. For example:

  1. Inspect the visual tree to understand the current UI state.
  2. Query for a specific button using a CSS selector.
  3. Tap the button to trigger an action.
  4. Take a screenshot to capture the resulting UI state.
  5. Assert that the expected changes occurred.

All of these steps are performed through MCP tool calls, enabling fully automated AI-driven testing and debugging.

Storage tools

The following tools enable AI agents to inspect and manage sandboxed app files on device or simulator.

Tool Purpose
maui_storage_roots List file storage roots advertised by the app. Call this before specifying a non-default root.
maui_files_list List files and directories under an advertised app storage root. Optionally specify a subdirectory path.
maui_files_download Download a file from an advertised app storage root. Returns the file content as base64.
maui_files_upload Upload a file to an advertised app storage root. Content must be base64-encoded. Parent directories are created automatically.
maui_files_delete Delete a file from an advertised app storage root.

Tool parameters

maui_storage_roots

Parameter Type Description
agentPort int? Agent HTTP port (optional if only one agent connected)

maui_files_list

Parameter Type Description
path string? Subdirectory path relative to the selected storage root (optional, lists root if omitted)
root string? Storage root id (default: appData; call maui_storage_roots to discover supported roots)
agentPort int? Agent HTTP port (optional if only one agent connected)

maui_files_download

Parameter Type Description
path string File path relative to the selected storage root
root string? Storage root id (default: appData)
agentPort int? Agent HTTP port (optional if only one agent connected)

maui_files_upload

Parameter Type Description
path string File path relative to the selected storage root
contentBase64 string File content encoded as base64
root string? Storage root id (default: appData)
agentPort int? Agent HTTP port (optional if only one agent connected)

maui_files_delete

Parameter Type Description
path string File path relative to the selected storage root
root string? Storage root id (default: appData)
agentPort int? Agent HTTP port (optional if only one agent connected)

See also

  • Last updated on