Welcome to MCPChecker quickstarts! These hands-on examples help you get started with testing your MCP servers using MCPChecker.
MCPChecker is a testing framework for MCP (Model Context Protocol) servers. It helps you verify that:
- Your tools are discoverable by AI agents
- Tool descriptions are clear and actionable
- Agents can correctly use your tools
- Your server handles edge cases properly
Think of it as integration testing for AI tool use.
Difficulty: Beginner Time: 5 minutes What you'll learn:
- Set up a basic MCP server
- Write your first MCPChecker test
- Run tests with Claude Code agent
- Understand test results
Perfect for: First-time users who want to understand the basics of MCPChecker.
Difficulty: Beginner Time: 10 minutes What you'll learn:
- Test a stdio/local MCP server (not HTTP)
- Configure MCPChecker for subprocess-based servers
- Write evals for system diagnostic tools
- Test read-only tools that gather system information
Perfect for: Users who want to test local CLI tools or understand stdio transport vs HTTP.
Key insight: Stdio servers are simpler for local-only tools - no HTTP endpoints needed, MCPChecker manages the subprocess lifecycle.
Difficulty: Intermediate Time: 15-20 minutes What you'll learn:
- How documentation quality affects agent success
- Compare bad vs good tool documentation (2 iterations)
- See identical code produce different test results
- Understand what makes tools discoverable when they overlap in functionality
Perfect for: Users who want to see MCPChecker's real value and learn documentation best practices.
Key insight: Same functionality (4 text processing tools), different documentation = different test results. Proves MCPChecker tests discoverability, not just functionality.
Have an idea for a quickstart? We'd love to hear from you! Check out our contributing guidelines to get started.