Add an external MCP server in Katalon AI Assistant

Last updated: August 2026

This document explains how to connect to an external MCP server in Katalon AI Assistant.

Supported MCP servers

Katalon Studio provides 3 built-in MCP servers that help you query Katalon's knowledge base, write/edit test cases in Katalon Studio, and interact with True Platform objects.

For external MCP servers, we support the followings:

Transport method	Execution environment	User type	Requirement
SSE (HTTP with Server-Sent Event)	Local or remote	Multiple users can access the MCP server	URL to the SSE endpoint
HTTP (Modern Streamable HTTP)	Local or remote	Multiple users can access the MCP servers	URL to the HTTP endpoint
STDIO	Local	Single user on the local machine running the MCP server	Server command and arguments

note

STDIO MCP servers are supported natively in Katalon AI Assistant, so you no longer need to use MCP proxy for these servers. If you are using an earlier version of Katalon Studio, set up MCP proxy as a workaround. Servers hosted via this method can be treated as SSE/HTTP types to connect to Katalon AI Assistant. Learn more at Set up MCP proxy for STDIO MCP servers below.

Add an external MCP Server

Follow these steps to add an MCP server:

1. Click the Settings icon, then select Agent Mode settings.
2. Click Add MCP Server:

3. Fill in these fields:

• Transport Type: HTTP or SSE. Your MCP server provider should specify this. Check out our external MCP servers for common servers you could add.
  • For HTTP and SSE: input the MCP server's URL.

• For STDIO: the command used to start the MCP server and its arguments. The syntax should be:

{
  "command": "npx",
  "args": ["-y", "chrome-devtools-mcp@latest"]
}

• Name: this name should contain no spaces or special characters (for example, "atlassian-mcp-server"). This name should be unique for the server list, and kept as short as possible (see limitation notes below).
• Title (Optional): the title is the display name of the server. You can have servers with the same name, but we highly recommend giving unique titles to servers.

4. Click OK. Once the server is added, you can view its tool list.

Known limitation

• AI providers (OpenAI, Azure OpenAI, Gemini, and AWS Bedrock) enforce a limitation of 64 characters on the combined length of the MCP Server name and tool name ({server_name}_ _ _{tool_name}). Exceeding this limit may cause Katalon AI Assistant Agent mode to fail when calling these APIs.

• To prevent this, try to keep the MCP Server name (or the tool's name if you host your own server) short, or temporarily disable the tool and revert back later. The Title of the MCP server is not affected by this limit, and can remain as long as needed.

• Some AI service providers also enforce a limit of 128 MCP tools per request. Starting in Katalon Studio 11.3.0, Katalon AI Assistant can automatically select the tools most relevant to your prompt before an Agent mode request. This helps reduce the chance of hitting the tool limit when you add external MCP servers. If the error still occurs, disable tools or servers you do not need.

Authentication for MCP servers

Some servers (like Atlassian) require authentication:

To authenticate, click Authorize. You'll be redirected to the server's authentication site.

When a server's token expires, Katalon Studio reminds you to re-grant authorization.

Simply click Authorize, and follow the steps above again. If it doesn't load properly due to connection issues, you can reload it with the Reload button.

Servers protected by Microsoft Entra ID

Starting from 11.5.0, Katalon AI Assistant can connect to MCP server protected by Microsoft Entra ID - including servers your organization hosts internally, which don't appear in the list of external servers below. If the server accepts a Microsoft work or school account, you can add it. See Set up Azure DevOps MCP server.

A few things to know about these servers:

• They reconnect on their own. The next time you open Katalon Studio, the server connects without asking you to sign in again.
• Each server signs in separately. Two servers can use two different accounts, and signing out of one doesn't affect the other.
• Removing a server also removes its sign-in. If you add the server back, you'll be asked to choose an account again.

If your organization's server is behind a company proxy, configure the proxy in Katalon Studio first. See Configure proxy preferences in Katalon Studio.

Information for your Entra administrator

Before users can sign in, your Microsoft Entra administrator may need to grant consent for Katalon Studio, or allow-list it on the server that issues tokens. Use these values:

Field	Value
Application (client) ID	aaf89ad7-776c-4054-97b2-828ebf82be91
Publisher	Katalon, Inc.
Application type	Public client (mobile and desktop)
Redirect URI	http://localhost
Tenancy	Multi-tenant

This client ID is the same in every installation of Katalon Studio. It doesn't vary by tenant, region, or environment, and it can't be changed through configuration.

Katalon Studio requests only the scope the MCP server advertises, plus the standard openid, profile, and offline_access. It requests no permissions of its own.

External MCP servers to add to Katalon AI Assistant

You can easily add any of the available open MCP servers from the remote MCP servers list.

See below some of the OAuth MCP servers we currently support, grouped by purposes.

Project management

Integrate with these MCP servers to enable Katalon AI Assistant to:

• Create and update ticket descriptions automatically
• Read requirements, documentation, and test plans as context for test generation
• Log and create bugs when tests fail

Server	Transport type	Server URL/Command and Arguments
Atlassian	HTTP	https://mcp.atlassian.com/v1/mcp
Notion	HTTP	https://mcp.notion.com/mcp
	SSE	https://mcp.notion.com/sse
Linear	HTTP	https://mcp.linear.app/mcp
Asana	SSE	https://mcp.asana.com/sse
Azure DevOps MCP (ADO)	HTTP when using mcp-proxy	See further setup instructions at Setting up MCP server for Azure DevOps.
	STDIO (KS 11.1.0+)	{ "command": "npx", "args": ["-y", "@azure-devops/mcp@next", "Katalon"] }
GitHub	HTTP	https://api.githubcopilot.com/mcp/ with PAT
	STDIO	{ "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"], "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxxxxxxxxxxx" } }
GitLab	STDIO (KS 11.1.0+)	Requires Node.js 18+ (so npx works) and a GitLab personal access token with the api scope. { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-gitlab"], "env": { "PATH": "/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin", "GITLAB_PERSONAL_ACCESS_TOKEN": "<your-token>", "GITLAB_API_URL": "https://gitlab.com/api/v4" } } Note about PATH: On macOS, keep the PATH property. Run echo $PATH in Terminal, then copy the result and use it as the PATH value. On Windows, remove the entire "PATH": "...", property from the configuration.
Google Drive	STDIO (KS 11.1.0+)	{ "command": "npx", "args": ["-y", "@piotr-agier/google-drive-mcp"] }. Requires GCP OAuth setup — see further setup instructions at Set up Google Drive MCP server.

See how we use Katalon AI Assistant to create a test case from requirements, with Atlassian MCP server:

Testing and automation

Integrate with these MCP servers so Katalon AI Assistant can directly interact with browsers and automation tools — enabling test case generation, execution, and debugging.

note

• Starting Katalon Studio version 11.0.0, you can use Katalon WebUI MCP tools to interact with your AUT. See Katalon WebUI MCP tools for more details.

Server	Transport type	Server URL/Command and Arguments
Selenium	HTTP when using mcp-proxy	1. Run: mcp-proxy --port 3000 --host 127.0.0.1 --stateless --allow-origin "*" -- npx -y @angiejones/mcp-selenium 2. The server will run locally at http://127.0.0.1:3000/mcp - use this URL to set up connection in Katalon AI Assistant.
	STDIO (KS 11.1.0+)	See angiejones/mcp-selenium for setup and getting the command/arguments.
Playwright	HTTP when using mcp-proxy	1. Run: mcp-proxy --port 3000 --host 127.0.0.1 --stateless --allow-origin "*" -- npx -y @playwright/mcp@latest 2. The server will run locally at http://127.0.0.1:3000/mcp - use this URL to set up connection in Katalon AI Assistant.
Chrome DevTools	HTTP when using mcp-proxy	1. Run: mcp-proxy --transport streamablehttp --port 8080 -- npx -y chrome-devtools-mcp@latest 2. The server will run locally at http://127.0.0.1:8080/mcp - use this URL to set up connection in Katalon AI Assistant.
mobile-mcp	STDIO (KS 11.1.0+)	{ "command": "npx", "args": ["-y", "@mobilenext/mobile-mcp@latest"] }
Windows Desktop MCP	STDIO (KS 11.1.0+)	See CursorTouch/Windows-MCP for setup and getting the command/arguments. Note: Windows only. macOS is not supported.
Excel	STDIO (KS 11.1.0+)	See negokaz/excel-mcp-server for setup and getting the command/arguments.

See how we use Katalon AI Assistant to create a test case using Playwright MCP server and requirements pulled using Atlassian MCP server:

Design and prototyping

Connect to these MCP servers to allow Katalon AI Assistant to analyze design structures and use design assets as context for generating more accurate and visually aligned test cases.

Server	Transport type	Server URL/Command and Arguments
Figma	HTTP	Figma's remote MCP Server URL is currently not supported. As an alternative, you can host the MCP server on your local machine, and connect to this server locally: 1. Open a Figma Design file (latest Figma version). 2. In the toolbar at the bottom, toggle to Dev Mode. 3. In the MCP server section of the inspect panel, click Enable desktop MCP server. A confirmation message will appear once the server is running. 4. The server runs locally at http://127.0.0.1:3845/mcp. Use this URL to set up connection in Katalon AI Assistant.
Webflow	SSE	https://mcp.webflow.com/sse

Please note that some OAuth servers are not yet supported. We plan to include support for these in future updates. Servers that use Microsoft Entra ID are supported from Katalon Studio 11.5.0 - see Servers protected by Microsoft Entra ID.

Set up MCP proxy for STDIO MCP servers

note

• Starting Katalon Studio version 11.1.0, STDIO MCP servers are officially supported. You only need MCP proxy for STDIO MCP servers if you are using an earlier version of Katalon Studio.

MCP proxy is a tool that acts as an intermediate and handles communication between an LLM model and the target MCP server. Once you set up MCP proxy, you can host multiple MCP servers with STDIO transport type on your local machine.

requirements

• Node.js (version 14 or higher) - Download here. npx is included automatically with Node.js
• Python - Download here.
• Either UV (installation guide here) or pipx (installation guide here) to install mcp-proxy

Run the following command in your terminal to install mcp-proxy. You can choose either UV:

uv tool install mcp-proxy

or pipx:

pipx install mcp-proxy

Once the MCP proxy is installed, you can host an MCP server with a command line (replace {{ package }} with the MCP server's package, e.g. @angiejones/mcp-selenium for Selenium's MCP server)

mcp-proxy --port 3000 --host 127.0.0.1 --stateless --allow-origin "*" -- npx -y {{ package }}

Customize the parameters if you wish:

Parameter	Description
--port 3000	Server port (customize as needed)
--host 127.0.0.1	Server host (localhost)
--stateless	Enable stateless mode
--allow-origin '*'	Allow all CORS origins (use specific origin in production)