The mabl MCP server connects your AI client to mabl’s testing tools so you can spend less time on repetitive tasks and free up more time for other work. Use natural language prompts to automate day-to-day tasks directly from your IDE or preferred workspace, including but not limited to:
- Running tests
- Getting results and failure analyses
- Creating new tests based on prompts
- Integrating with the other MCP tools in your workflow
Interested in trying it out? Follow these steps to get started.
Prerequisite
To configure and use the mabl MCP, you must have an AI client installed on your system that supports MCP servers.
Setup
The mabl MCP is compatible with any AI client that supports MCP servers. For specific instructions on MCP setup, consult your AI client’s documentation. The following tabbed section provides guidance on how to configure the mabl MCP for specific clients.
Config
To add the mabl MCP server in Claude Desktop, open Claude Desktop Settings from your system’s menu bar: Claude > Settings…
Accessing settings in Claude
In the Settings window, click on the
Developer tab. Click on Edit Config
to open the configuration file for Claude Desktop. This action creates
or opens the existing claude_desktop_config.json file. Depending
on your OS, this file is located in one of the following locations:
-
macOS:
~/Library/Application Support/Claude/claude_desktop_config.json -
Windows:
%APPDATA%Claudeclaude_desktop_config.json
Add the following JSON to configure the mabl MCP server.
{
"mcpServers": {
"mabl": {
"command": "npx",
"args": ["@mablhq/mabl-cli@latest", "mcp", "start"],
"env": {}
}
}
}Start the server
To confirm the mabl MCP server is installed, quit and restart Claude Desktop. When you click on the settings icon at the bottom of the conversation input box, you should see available mabl MCP tools.
Viewing mabl MCP tools in Claude Desktop
Troubleshooting
If setup wasn’t successful, follow these steps to troubleshoot:
Check Claude Desktop errors
On startup, Claude surfaces errors as toast messages. Click on the toast message to review the error and access logs from Developer Settings. Alternatively, you can access it from the Claude Desktop menu bar: Settings > Developer.
Look for error messages containing “mabl” or “MCP”. Common issues include:
- “Command not found” or “No such file or directory”
- “Permission denied”
- Connection timeout errors
Verify your configuration
In your claude_desktop_config.json file, confirm that the
path exists and is executable.
Check for conflicting Node versions
If you have multiple versions of Node installed, Claude may be choosing an unsupported version when it starts up the mabl MCP server. In this situation, try uninstalling all versions of Node and re-installing an LTS version.
For more details on managing MCP servers in Claude Desktop, refer to their official documentation.
Before you start, make sure GitHub Copilot is enabled and you are signed into your GitHub account in VS Code.
Config
To install the mabl MCP in VS Code, visit the setup page and select one of the following options:
- Install - install the mabl MCP for your user profile in VS Code, so that it is accessible across all projects on your local machine.
- Install in Workspace - configure the mabl MCP for a specific workspace, such as a shared project.
Alternatively, if you prefer to manually install the mabl MCP, use the following config block:
{
"inputs": [],
"servers": {
"mabl": {
"type": "stdio",
"command": "npx",
"args": [
"@mablhq/mabl-cli@latest",
"mcp",
"start"
]
}
}
}Start the server
To start the mabl MCP server, click on the start button within the “servers” object.
Make sure the GitHub Copilot chat is set to "Agent" mode before you start using the mabl MCP.
For more details on managing MCP Servers in VS Code, refer to their official documentation.
Config
To install the mabl MCP for your user profile in Cursor, visit the setup page and click on the Install button.
Alternatively, if you want to configure the mabl MCP for a specific project,
such as a shared project, create a file named
.cursor/mcp.json in the root of the project directory and
add the following config block:
{
"mcpServers": {
"mabl": {
"command": "npx",
"args": [
"@mablhq/mabl-cli@latest",
"mcp",
"start"
]
}
}
}After saving the file, Cursor should automatically detect the mabl MCP, and you can enable it.
For more details on managing MCP servers in Cursor, refer to their official documentation.
Config
You can also connect from any AI client that supports MCP servers. Follow the client’s MCP documentation for setup, and use the following configuration when adding the mab MCP server:
{
"mcpServers": {
"mabl": {
"command": "npx",
"args": ["@mablhq/mabl-cli@latest", "mcp", "start"],
"env": {}
}
}
}Authentication
Before you start using the mabl MCP, you need to be logged in with your user credentials in the mabl CLI. If you aren’t already authenticated in the mabl CLI, your AI client will walk you through how to do that.
Alternatively, you can manually authenticate from the terminal with the command mabl auth login.
Usage
With MCPs, you can carry out your tasks in a back-and-forth dialogue instead of using technical commands. In the AI chat, start sending natural language prompts to interact with the mabl MCP, such as:
- “Is there a test in my mabl workspace that validates…”
- “Get the latest results for that test.”
- “Can you run that test locally? Use this URL…”
Check out the article on mabl MCP tools and prompts for a full list of supported actions.
Prompting
Writing effective prompts is the key to getting the most out of the mabl MCP. Check out the article on how to interact with the mabl MCP for best practices.
To demonstrate how you might use the mabl MCP, consider the following sample prompts.
Run a test locally
User: “Do I have a mabl test that validates the login?”
AI: Responds with a list of mabl tests.
User: “OK, let’s run the first test locally using my local environment.”
AI: Gets the local environment and runs the test using the mabl CLI.
Create a new mabl test after a code change
User makes some code change
User: “How would you test this using an E2E tool? Give me an intent and a list of steps.”
AI: Evaluates the change and provides a list
User: “Perfect, let’s create a mabl test with that info. Use the URL generated in the terminal.”
Get deployment results
User: “What happened with my previous mabl deployment?”
AI: Gets the last commit and queries deployments in mabl with that commit.
Integrate with the Atlassian MCP Tool:
User: “I want to start working on the jira ticket SCRUM-005”
AI: Returns some information about the jira ticket
User: “Let’s implement that change”
AI: Makes the change
User: /mcp.mcp-mabl.create-mabl-test-for-current-work
AI: Launches the mabl test creation based on the code change