Connect an AI assistant with MCP

The TestCollab MCP server gives an assistant tools for cases, suites, plans and project context. Version 1.4 adds the tools to read a plan’s executable steps and report results, step outcomes, time and evidence. Requests use the TestCollab permissions of the configured user.

Configure a local MCP server

Use Node.js 22 or newer. In a client that supports command-based MCP servers, add an entry like this:

MCP server configuration

{
  "mcpServers": {
    "testcollab": {
      "command": "npx",
      "args": ["-y", "@testcollab/[email protected]"],
      "env": {
        "TC_API_TOKEN": "your-api-token",
        "TC_API_URL": "https://api.testcollab.io",
        "TC_DEFAULT_PROJECT": "123"
      }
    }
  }
}

Use your client’s secret storage where available. Do not commit a configuration containing a real token.

For an EU account, change TC_API_URL to https://api-eu.testcollab.io. TC_DEFAULT_PROJECT sets the project used by your prompts. If you work across projects, name the project in your request.

Verify the connection

Restart or reload the client, list the server’s tools, and start with a read:

Show my suites and test cases related to checkout.

Ask the assistant to inspect project context before creating or editing records, so it uses the existing suite, tag, field and member IDs.

What to use it for

• Find relevant test cases and inspect their steps.
• Prepare a test plan from an explicit case selection.
• Run an entire test plan with your assistant’s browser or test tools, then report results and evidence through MCP.
• Organize suites and review proposed changes.
• Read a plan’s runs, configuration and progress.

Start with run an entire test plan, or explore the other MCP workflows for concrete prompts and tool names.