Skip to main content

​
Overview

Spur’s MCP (Model Context Protocol) server connects your AI assistant directly to your Spur tests. Whether you use Cursor, Claude Code, ChatGPT, or any MCP-compatible client, you can discover tests, trigger runs, and debug failures through natural language – without leaving your development environment. Authentication uses OAuth. When you connect for the first time, your browser will open to authorize access to your Spur account.

​
How to Setup

​
Cursor

Click this link to install! Install MCP Server

​
Claude Code

Run this command in your terminal: 
claude mcp add Spur -- npx -y mcp-remote@latest https://app.spurtest.com/api/mcp

​
ChatGPT

  1. Enable Dev mode in your workspace settings (requires owner permissions)
  2. Create a new App
  3. Set the MCP URL to https://app.spurtest.com/api/mcp
  4. Set Auth to OAuth

​
Available Tools

list_tests – Lists all tests in the active application with IDs, suites, and descriptionsget_test_details – Returns the full definition of a test: steps, environments, viewports, and scenario rows
run_test – Runs a single test by test_id and env_id. Optionally specify browser, viewport, or scenario rowrun_tests – Runs multiple tests in one collection run, grouped by shared configuration
get_test_run_overview – Start here. Summarizes a run’s status, step results, warnings, and failuresget_test_run_details – Deep dive into steps, sub-steps, configs, and artifacts. Use after the overviewget_test_runs – Lists the last 50 runs for a given test
get_test_run_console_logs – Browser console output and JavaScript errors (last 100 entries)get_test_run_network_logs – HTTP requests and responses captured during the run (last 5 entries)get_test_run_screenshots – Screenshots from the test execution for visual inspection
list_applications – Lists all applications on your account and shows the active oneswitch_application – Switches the active application for subsequent calls

​
Common Workflows

​
Discover and Run Tests

Ask your assistant: “What tests do I have for checkout?” The agent calls list_tests, scans names and descriptions, and returns matching tests. You can then run specific tests directly from the conversation. Show Image Natural language test discovery through your IDE

​
Execute with Specific Configurations

Ask: “Run the checkout test on staging with mobile viewport” The agent calls run_test with the correct environment and viewport, then monitors the job status.

​
Debug Failed Runs

Ask: “What went wrong in this run?” The agent calls get_test_run_overview to identify failing steps, then drills into get_test_run_details, get_test_run_console_logs, or get_test_run_screenshots to surface the root cause. Show Image Debugging workflow from overview to console logs

​
Best Practices

  • Start with the overview: When analyzing results, always use get_test_run_overview first. This provides context before diving into detailed logs or screenshots.
  • Be specific in your prompts: “Run the checkout test on production” works better than “run some tests.” Include environment names, viewport preferences, or scenario details when relevant.
  • Model quality matters: More capable models (Claude Opus, GPT-4) are better at choosing the right tools and interpreting results. Smaller models may need more explicit guidance.
  • Use multi-test runs strategically: When testing the same flow across environments, use run_tests to group configurations efficiently rather than running tests individually.