MCP integration

This section describes how to control OpenText Enterprise Performance Engineering through external AI tools using the Model Context Protocol (MCP).

About MCP

The Model Context Protocol (MCP) is an open standard that enables AI assistants to securely connect with external tools and data sources. When you make a natural language request, such as "Run the demo script test in project XYZ", the MCP client translates this into a standardized protocol format and sends it to the MCP server. The server applies security checks, performs out the requested operation, and returns the result. This seamless flow transforms your conversational requests into actual operations.

This natural language approach eliminates the need to remember specific IDs, navigate multiple UI screens, or write API scripts. Just describe what you need, and the AI tool handles the technical details.

Back to top

MCP server Capabilities

The MCP integration supports over 20 tools, enabling comprehensive test management and monitoring through natural language.

The supported tools provide the following capabilities.

Area Capabilities
Project management

  • Get current domain and user configured in the system

  • Get all domains available to the user

  • Get all projects on current domain available to the user

  • Change current domain and project
Test configuration

  • Get all tests in the project

  • Get all test sets

  • Get all scripts in the project

  • Get all test folders in the project

  • Get load test details

  • Check test validity
Test execution

  • Start new test run

  • Stop test run

Note:

  • Only basic workload by test by number can be defined.

  • Automatch load generators will be used.

  • Default runtime settings for the protocol are defined.

  • Vusers will start and stop simultaneously.
Results & monitoring

  • Get all runs of a specific test

  • Get active runs of a specific test

  • Get test run status, get test run results

  • Get online summary data

  • Get extended run data

  • Download test run results

  • Start late analysis

  • Post-run action parameter can be defined during Start Run/Stop run
Host management Get list of hosts available in the project

Each tool corresponds to a single public API operation, making it easy to chain operations together for complex workflows.

Back to top

Integrate with the MCP server

This section explains how to integrate with the MCP server.

To integrate with the MCP server:

  1. Prerequisite. An external AI tool must be installed and configured. For example, Claude Desktop application.

  2. Install the MCP server.

    1. Copy the MCP server installer from <server_Installdir>\Additional Components\MCPServer to the local machine where your AI tool is installed.

      Note: MCP Server is also available from Download Applications in the Performance testing application banner. For details, see Download applications and components.

    2. Run the installer to complete the MCP server setup.

  3. Configure the connection.

    1. In the MCP Server Connection Configuration window, provide the following OpenText Enterprise Performance Engineering details:

      • Server URL

      • Client ID

      • Client Secret

      • Domain

      • Project

      • (Optional) Tenant ID

    2. Click Save to confirm connectivity and close the window.

  4. Update the AI tool configuration.

    Add the MCP server to the tool’s MCP configuration file (commonly mcp.json; some tools use alternatives such as claude_desktop_config.json). The file name and location depend on the tool.

    Example structure to add:

    Copy code 
    {
      "mcpServers": {
        "OT Enterprise Performance Engineering MCPServer": {
          "command": "C:\\{MCP_Server_Installation_Root}\\EnterprisePerformanceEngineering_MCPServer.exe",
          "args": ["--silent"]
        }
      }
    }

  5. Registration.

    The MCP server registers automatically with the external AI tool and exposes OpenText Enterprise Performance Engineering tools.

    Note: All connection-related data is stored encrypted.

Back to top

Network security configuration

To mitigate Server-Side Request Forgery (SSRF) risks, the MCP server includes a NetworkSecurity section in its configuration.

Use this section to control which hosts can be accessed and enforce security rules, while allowing flexibility for internal or development scenarios.

Key settings Details
AllowedHosts

Define the host names that the MCP server can call (exact names and wild cards are supported).

Recommended: Restrict access to your hosts to minimize exposure.
AllowHttp

Controls whether plain HTTP is permitted. If set to true, both HTTP and HTTPS are allowed.

Recommended: Set to false (the default value) so that only HTTPS is allowed.
AllowPrivateIPs

Blocks or permits private/internal IP ranges and IPv6 internal ranges.

Recommended: Set to falseto block private networks unless they are explicitly listed in AllowedHosts.

Back to top

URL validation rules

The following rules ensure that all server URLs are properly validated for security and connectivity before being used by the MCP server.

Area Details
General URLs must be valid and absolute.
Scheme restrictions

  • By default, only HTTPS URLs are allowed.

  • If AllowHttp = true, both HTTP and HTTPS are allowed.

  • Any other scheme, such as FTP or file, is rejected.
Host validation The host must exist and be reachable using DNS.
Private IP restrictions

  • Private IPs are blocked unless explicitly allowed.

    Example of private IPs: 10.x.x.x, 192.168.x.x, 172.16–31.x.x.

  • Private IPs are allowed only if:

    • Listed in AllowedHosts, or

    • AllowPrivateIPs = true

  • A private IP explicitly listed in AllowedHosts overrides AllowPrivateIPs = false.
Allowed hosts

  • Only explicitly allowed hosts or "*" for public hosts are accepted (public hosts are not recommended for production).

  • Wildcards (*.example.com) are supported for subdomains.

  • Host names in AllowedHosts must not include protocol prefixes.

  • If a host is not allowed and not covered by "*", it is blocked.

Back to top

Usage examples

The following are examples of creating, running, and managing tests using natural language.

Example 1: Create a Test

Example Command Steps
“Create a new performance test.”

  1. The AI tool shows the available scripts and hosts.

  2. Select a folder for the test.

  3. Provide a test set name. If a test set does not exist, a new one is created.

  4. Select the workload type: Basic workload only, by test or by number.

  5. Define the groups (scripts) to use and enter the number of Vusers to allocate to each script. For example, "10 Vusers for every script in the folder”.

  6. Define the duration of the scheduler for the entire test. This includes Vuser initialization, start, and stop time. For example 30m or 1h.

  7. Assign Automatch Controller and load generators.

  8. Choose load generator placement: per test or per group.

  9. The test is created and ready to run.

Example 2: Run a Test

Example Commands Steps

  • “Show me all tests in the project.”

  • “Show me details of test 1011.”

  • “Run test Checkout.”

  1. The AI tool lists the available tests (Name, ID, Location).

  2. Select a test by name or ID.

  3. The AI tool validates the test.

  4. Choose a test set from the list.

  5. Choose a post-run action, such as download report.

  6. The test starts and the AI tool returns the Run ID.

Example 3: Manage test runs

Example Commands Capabilities

  • “List runs for Checkout.”

  • “Show active runs.”

  • “Stop run 9722.”

  • View the list of runs for a specific test, including run ID, state, and start time.

  • View details of active runs, including run ID, test name, start time, and duration.

  • Stop an active run by providing the run ID. If you don't provide the run ID, you are prompted to choose a post-run action.

  • Start late analysis.

  • Get online summary or extended run data.

  • Download results. For example, reports.zip.

Back to top