Skip to content
MCP server

MCP Test Utils

Windows desktop UI automation and regression testing for AI agents via MCP

Runs locallyStatus and history
Install
Pricing
Free
Platforms
Windows, CLI
Runs
Locally
Version
3.27.23
Listing updated
today

Similar MCP servers

About MCP Test Utils

MCP Test Utils is a free MCP server for developer tools. It runs locally on your machine and is started by your AI client.

The README says this server gives an AI agent screenshots, window control, mouse and keyboard input, UI Automation trees, and OCR, so it can drive a Windows application the way a tester would. Each click can be verified against the expected effect and retried, and a single binary runs with no runtime or installer.

Beyond interactive use, saved scenarios can be replayed without an agent from a script or CI job, with exit codes, JUnit output and per-row data runs. A recording mode turns a manual walkthrough into a draft scenario. Process metrics such as memory, handles and GDI objects can be sampled repeatedly to spot leaks.

The project targets QA engineers and developers testing Windows desktop software. It is currently Windows x64 only, and scenario runs are limited to 50 steps unless a license key is registered, while sealed scenarios run without the limit.

Install

Run in a terminal

claude mcp add mcp-test-utils -- npx -y https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe

Add to claude_desktop_config.json

{
  "mcpServers": {
    "mcp-test-utils": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "mcp-test-utils": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "mcp-test-utils": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "mcp-test-utils": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "mcp-test-utils": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Configuration

  • MCP_LOG_DIRDirectory for logging sessions (opt-in). Without it, logging tools are hidden.
  • MCP_LOG_MAX_MBSession size limit in MB. Warning on exceed.
  • MCP_LOG_RETAIN_DAYSAuto-delete log sessions older than N days. 0 to disable.
  • MCP_SEARCH_DIRSAllowed directories for search_in_files (semicolon-separated). Without it, the tool is hidden.
  • MCP_LAUNCH_APPSPrograms (exe files or folders, semicolon-separated) that launch_app may start. Without it, launch_app starts nothing.
  • MCP_NOTIFYSet to off to hide the tray icon and the live call log window.
  • MCP_INPUT_LOCKSet to off to disable the keyboard and mouse lock while the server sends input (on by default; Esc stops).
  • MCP_INPUT_LOCK_STALL_MSSilence of the server in ms after which the input lock lets the input go.

Tools this server provides (31)

Taken from the project documentation; the running server may differ.

take_screenshot
Screenshot of the entire desktop with quality presets.
take_window_screenshot
Screenshot of a specific window.
read_screen_text
OCR the entire screen.
read_region_text
OCR a screen region with word coordinates.
list_windows
List windows with id, title, app, process id, position, size and state.
focus_window
Bring a window to the front and optionally set its position and size.
launch_app
Start a program from the configured allowlist and wait for its window.
close_window
Close a window as its X button would and report save prompts.
mouse_click
Click an element or coordinates with optional wait and retry.
mouse_move
Move the cursor to a point or UI element.
mouse_drag
Drag between elements or points at a human pace.
mouse_scroll
Scroll the mouse wheel over an element.
keyboard_type
Type Unicode text and verify the field shows it.
keyboard_press
Press a key, optionally several times.
keyboard_shortcut
Press a key combination.
list_ui_elements
Return the UI Automation tree with filters.
get_focused_element
Report which element has keyboard focus.
read_text
Read full text of a document or field via UI Automation.
ui_action
Invoke, toggle, expand, select, set value or scroll an element.
check_accessibility
Report accessibility defects in controls.
process_info
Report memory, handles, GDI objects, threads and CPU of the app.
wait_for
Wait for a window, element or state.
batch
Run up to 20 tool calls in one request.
run_scenario
Replay a saved scenario, suite or data rows and get a report.
record_scenario
Record actions in one window into a draft scenario.
search_in_files
Search text or regex in allowed directories.
find_files
Find files by glob pattern in allowed directories.
get_usage_guide
Usage guide for agents with sections on request.
prepare_feedback
Prepare a suggestions file for the author.
enable_logging
Start recording tool calls to JSONL and screenshots.
disable_logging
Stop recording and get session stats.

What it does

Verified UI interaction
Clicks, typing and drags can wait for the expected effect and retry, and typed text is checked against the field.
UI Automation access
Reads the element tree with ids, values, states and coordinates, and acts on elements without mouse or keyboard.
Scenario regression runs
Replays saved scenarios, suites and CSV data rows with pass/fail reports and JUnit XML for CI.
Recording into drafts
Captures actions in one window into a draft scenario that an agent can later refine.
Accessibility checks
Flags unnamed controls, keyboard-unreachable controls, duplicate ids and zero-size controls.
Resource monitoring
Reports memory, handles, GDI objects, threads and CPU of the application under test.

Good for

  • Run an agent-driven QA pass through an application and write a test report
  • Replay saved desktop scenarios after each build in a CI job
  • Record a manual walkthrough and turn it into an automated regression test
  • Audit a Windows app for accessibility defects in its controls
  • Check a desktop application for handle or GDI leaks across repeated actions

Pricing

Free

All tools free; a license key lifts the 50-step scenario limit.

MCP Test Utils FAQ

Which operating systems are supported?

The README says it is Windows x64 only, with no macOS or Linux builds yet.

Is it free to use?

The README says all tools are free; a license key only lifts the 50-step limit of scenario runs.

How do I get a license key?

Keys are issued through early access; the README says the first 30 users get a free one-year key in exchange for feedback, requested by email with a machine code.

Does it send data over the network?

The README says it opens no network ports and sends no data externally, going online only when the user clicks Check for updates.

Is MCP Test Utils free?

Yes. MCP Test Utils is free to use.

How do I add MCP Test Utils to Claude Code?

Run this command in your terminal: claude mcp add mcp-test-utils -- npx -y https://github.com/JeenyJAI/mcp-test-utils/releases/download/v3.27.23/mcp-test-utils.exe

What are the best alternatives to MCP Test Utils?

The closest alternatives to MCP Test Utils are bad-ass-mcp, Mobai MCP, MCP Server, simplify3x-connect and Chrome DevTools MCP.

Where this comes from

  • Description: written from the project's README and its MCP Registry entry.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers