mcp-probe-kit

mcp-probe-kit is a comprehensive MCP server that provides 21 tools spanning the entire development lifecycle, from product analysis to final release. It emphasizes structured, machine-readable output to facilitate tool chaining and state tracking, with both orchestration and atomic tools designed to return JSON metadata, plans, and artifacts. Core orchestration tools like start_feature, start_bugfix, start_onboard, start_ui, start_product, and start_ralph enable one-click, multi-tool workflows that automatically coordinate analysis, design, estimation, and implementation steps. The toolkit is compatible with major MCP clients and includes a delegated orchestration flow where AI calls tools step-by-step and persists files, guided by a structured plan in structuredContent.metadata.plan. When you run the server, you gain access to all 21 tools (e.g., code_review, fix_bug, gentest, init_project, add_feature, ui_design_system, sync_ui_data, and more) and their ability to emit structured outputs that can be chained into a larger automation pipeline.

Prerequisites:

• Node.js (v18 or newer) and npm installed on your system
• Basic familiarity with running MCP servers via npx or package runners

Installation steps:

1. Ensure Node.js and npm are installed. Verify with: node -v npm -v

2. Install or run the MCP server using NPX (recommended for quick start): npx -y mcp-probe-kit

3. Alternatively install globally (optional): npm install -g mcp-probe-kit Then run with: mcp-probe-kit

4. If you need to install via package manager and pin a version, you can also run: npm i -g mcp-probe-kit@<version>

5. Verify the server is running by checking the console output or visiting the configured MCP client connection endpoint as documented in the project docs.

Tips and common issues:

• Use the delegated orchestration mode by default; ensure your client supports the plan-based workflow and that you persist outputs as the plan dictates.
• The server outputs structuredContent with artifacts and metadata.plan for start_* tools; use these to chain tools in an automated pipeline.
• If you encounter permission errors with NPX, try using a local npm cache or run with elevated permissions where appropriate.
• Keep your environment variables in sync with your deployment environment if using CI/CD; typical vars are not strictly required for local runs but may be necessary in production.
• Refer to the 21-tool list in the docs to understand which tools are available and how to compose multi-step workflows (e.g., init_project, add_feature, interview, ui_design_system, gentest).