Devpipe MCP Server

# How to Prompt the Devpipe MCP Server This guide explains how to effectively use natural language to interact with the devpipe MCP server through your AI assistant. ## 🎯 The Key Principle **Always specify which project you want to work with in your prompt.** The MCP server doesn't automatically know which project you're working on. You need to tell it by: 1. Referring to "this project" or "this repo" (uses your current workspace) 2. Using a project name that the AI can infer 3. Providing an explicit absolute path ## ✅ Good Prompts ### Using "this project" or "this repo" ``` "Analyze this project" "Run devpipe in this repo" "Show me the tasks in this project" "What technologies are in this repo?" ``` ### Using Project Names ``` "Analyze the devpipe project" "Run tests in my go-app" "List tasks from the people project" ``` ### Using Explicit Paths ``` "Analyze /Users/you/projects/my-app" "Run devpipe in /Users/you/projects/go-service" "Show tasks from /Users/you/projects/web-app/config.toml" "What's in /Users/you/repos/devpipe?" ``` ### Working with Multiple Projects ``` "Analyze /Users/you/projects/project-a" "Now run devpipe in /Users/you/projects/project-b" "Compare tasks between this project and /Users/you/projects/other-project" ``` ## ❌ Vague Prompts (May Not Work) ``` "Analyze the project" # Which project? "Run devpipe" # Where? "Show me the tasks" # From which config? ``` **Note:** These might work if the AI can infer context from your workspace or conversation history, but it's better to be explicit. ## 📋 Common Patterns ### Analyzing Projects ``` "Analyze this project" "What technologies are in /Users/you/projects/new-app?" "Detect technologies in this repo" "Scan /path/to/project for devpipe tasks" ``` ### Listing Tasks ``` "Show me all tasks in this project" "List tasks from /Users/you/projects/my-app/config.toml" "What tasks are configured in the devpipe project?" ``` ### Running Pipelines ``` "Run devpipe in this project" "Execute the pipeline for /Users/you/projects/my-app" "Run only lint and test tasks in this repo" "Run devpipe with --fast in /Users/you/projects/go-service" ``` ### Debugging ``` "Why did the lint task fail in this project?" "Show me logs from /Users/you/projects/my-app" "What went wrong in the last run?" ``` ### Creating Configs ``` "Create a devpipe config for this project" "Generate a config for /Users/you/projects/new-app" "Bootstrap devpipe in this repo" ``` ### Generating Tasks ``` "Create a Go linting task" "Generate a Python unit test task" "Help me add a TypeScript type check" ``` ## 🔧 How It Works When you say: > "Analyze the project at /Users/you/repos/devpipe" The AI assistant: 1. Understands you want to analyze a project 2. Extracts the path `/Users/you/repos/devpipe` 3. Calls the MCP tool `analyze_project` with `projectPath: "/Users/you/repos/devpipe"` 4. Returns the results in a readable format **You never need to know:** - Tool names (`analyze_project`, `list_tasks`, etc.) - Parameter names (`projectPath`, `config`, etc.) - JSON syntax Just describe what you want in plain English! ## 💡 Pro Tips 1. **Be specific about the project** - Always mention which project you're working with 2. **Use absolute paths for clarity** - `/Users/you/projects/my-app` is clearer than "my app" 3. **"This project" works great** - When working in a single workspace 4. **You can switch projects mid-conversation** - Just specify the new path 5. **No configuration needed** - No need to set `DEVPIPE_CWD` or other environment variables ## 🚫 What You Don't Need - ❌ Setting `DEVPIPE_CWD` environment variable - ❌ Configuring workspace-specific settings - ❌ Restarting your IDE when switching projects - ❌ Knowing MCP tool names or parameters - ❌ Writing JSON or structured commands ## 📚 Examples by Use Case ### Single Project Workflow ``` "Analyze this project" "Run devpipe" "Show me the tasks" "Why did the lint task fail?" ``` ### Multi-Project Workflow ``` "Analyze /Users/you/projects/frontend" "Now check /Users/you/projects/backend" "Run tests in /Users/you/projects/api" "Compare configs between frontend and backend projects" ``` ### New Project Setup ``` "Analyze /Users/you/projects/new-service" "What technologies did you detect?" "Create a devpipe config for it" "Generate a Go linting task" ``` ### CI/CD Generation ``` "Generate GitHub Actions for this project" "Create GitLab CI config for /Users/you/projects/my-app" ``` ## 🎓 Learning Curve - **Beginner:** "Run devpipe in /Users/you/projects/my-app" - **Intermediate:** "Run only lint and test tasks in this project" - **Advanced:** "Run devpipe with --fast and --fail-fast, skipping the security scan, in /Users/you/projects/go-service" All levels work - the AI understands them all!