MCP 服务器

Agile Planner MCP Server

Automatically generates structured agile backlogs including epics, features, and user stories from natural language descriptions within AI-powered IDEs. It streamlines project management by creating AI-optimized markdown files and directory structures to guide step-by-step implementation.

README

Agile Planner MCP Server (v1.7.3) - AI-Powered Agile Backlog Generator

Agile Planner MCP automatically generates complete agile backlogs (Epics, User Stories, MVP, iterations) or specific features from a simple description, directly within Windsurf, Cascade, or Cursor, with no technical skills required.

Latest improvements (v1.7.3):

Correction du mode MCP pour generateFeature: Amélioration robuste de l'extraction des user stories

Structure RULE 3 renforcée: Creation cohérente des dossiers epics/features/user-stories

Résolution du problème Notepad sur Windows: Normalisation des flux stderr/stdout en mode MCP

Logs de diagnostic détaillés: Identification plus facile des problèmes

Restructuration du projet: Organisation claire des fichiers de test et temporaires

Mise à jour des guides d'utilisation: Instructions complètes pour Windsurf, Claude et Cursor

See CHANGELOG.md for full details.

Previous improvements (v1.7.1):

Refonte complète de la documentation MCP: Documentation détaillée de l'architecture serveur MCP avec diagrammes Mermaid.

Réduction de la complexité cognitive: Refactorisation majeure des modules critiques (json-parser, mcp-router).

Amélioration de la robustesse: Meilleure gestion des erreurs et tests d'intégration E2E optimisés.

See CHANGELOG.md for details.

❌ Without Agile Planner MCP

Creating agile backlogs manually is time-consuming and error-prone:

❌ Hours spent writing user stories, acceptance criteria, and tasks
❌ Inconsistent formatting and structure across different projects
❌ No clear implementation guidance for AI coding assistants
❌ Manual prioritization and organization without strategic framework

✅ With Agile Planner MCP

Gestion d’erreur centralisée

Tous les retours d’erreur des fonctions generateBacklog et generateBacklogDirect sont désormais formatés par handleBacklogError pour garantir l’uniformité du JSON et la robustesse de l’audit.
Les exemples d’erreur affichent le format : { success: false, error: { message: ... } }

Agile Planner MCP generates complete, structured agile backlogs with precise AI-guided annotations in seconds:

✅ Complete backlog structure with epics, features, user stories, and orphan stories
✅ AI-optimized annotations that guide implementation step-by-step
✅ Progress tracking with task checkboxes and dependency management
✅ Centralized organization in a dedicated .agile-planner-backlog folder
✅ Intelligent feature organization that automatically associates features with relevant epics

📑 Documentation

This documentation has been reorganized for better navigation:

User Guides

Guide d'intégration MCP - Guide d'intégration avec Claude, Cursor et Windsurf IDE
Guide d'utilisation optimal - Guide d'utilisation détaillé
Guide de migration - Guide pour migrer depuis les versions précédentes

Developer Documentation

Développement - Guide de développement
Spécifications MCP - Spécification du protocole MCP
Problèmes connus - Liste des problèmes connus et dette technique
Plan de refactorisation - Plan détaillé de refactorisation du code
Plan de refactorisation des tests - Plan de correction des tests
Roadmap - Feuille de route des versions futures
Architecture MCP - Architecture complète du serveur MCP
Système de génération Markdown - Architecture du générateur markdown
Format du backlog - Spécification du format JSON de backlog

Helper Functions

createApiMessages(project) - Génère la paire de messages système/utilisateur pour l'IA. Le paramètre project peut être une chaîne de type "Nom: description" ou un objet { name, description }.

Note TDD : Les assertions sur les erreurs doivent vérifier le format unifié { success: false, error: { message: ... } }. Toute modification du format d’erreur nécessite la mise à jour des tests d’intégration.

Architecture Documentation

Design - Design général du projet
Format de backlog - Format du backlog généré
Diagramme de validation de backlog - Diagramme de validation
Compatibilité Multi-LLM - Compatibilité avec plusieurs LLMs

🚦 Setting up in Windsurf / Cascade / Cursor

Ask your administrator or technical team to add this MCP server to your workspace configuration:

Copy .env.example to .env and fill in your OPENAI_API_KEY or GROQ_API_KEY.

Option 1: Using a local installation

{
  "mcpServers": {
    "agile-planner": {
      "command": "node",
      "args": ["D:/path/to/agile-planner/server/index.js"],
      "env": {
        "MCP_EXECUTION": "true",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

Option 2: Using the NPM package

{
  "mcpServers": {
    "agile-planner": {
      "command": "npx",
      "args": ["agile-planner-mcp-server"],
      "env": {
        "MCP_EXECUTION": "true",
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

🧠 How It Works

Describe your project in plain English, providing as much detail as possible.

SaaS task management system for teams with Slack integration, 
mobile support, and GDPR compliance.

Agile Planner MCP processes your description through a robust validation pipeline:
- 🤖 Leverages OpenAI or Groq LLMs to generate the backlog structure
- 🧪 Validates the structure against a comprehensive JSON schema
- 🔍 Enhances features with acceptance criteria and tasks
- 📝 Organizes stories into epics and features
- 🏗️ Creates a complete directory structure with markdown files
Receive a fully structured agile backlog in seconds:

Structure du dossier généré

.agile-planner-backlog/
├── epics/
│   └── [epic-slug]/
│       ├── epic.md
│       └── features/
│           └── [feature-slug]/
│               ├── feature.md
│               └── user-stories/
│                   ├── [story-1].md
│                   └── [story-2].md
├── orphan-stories/
│   ├── [story-orpheline-1].md
│   └── [story-orpheline-2].md
└── backlog.json

Note : Les dossiers planning/mvp et planning/iterations sont supprimés. Toutes les user stories sont générées dans leur arborescence épics/features ou dans orphan-stories si elles ne sont rattachées à aucune feature/epic. Le fichier backlog.json ne contient plus de sections mvp ou iterations.

All files include AI-friendly instructions to guide implementation. See the examples folder for sample outputs.

Commands

Agile Planner MCP supports the following commands:

Generate a Complete Backlog

// In Windsurf or Cascade
mcp0_generateBacklog({
  projectName: "My Project",
  projectDescription: "A detailed description of the project...",
  outputPath: "optional/custom/path"
})

// CLI
npx agile-planner-mcp-server backlog "My Project" "A detailed description of the project..."

Generate a Specific Feature

// In Windsurf or Cascade
mcp0_generateFeature({
  featureDescription: "A detailed description of the feature to generate",
  storyCount: 3,  // Optional: number of user stories to generate (min: 3)
  businessValue: "High", // Optional: business value of this feature
  iterationName: "iteration-2", // Optional: target iteration (default: 'next')
  epicName: "Optional Epic Name", // Optional: specify an epic or let the system find/create one
  outputPath: "optional/custom/path" // Optional: custom output directory
})

// CLI
npx agile-planner-mcp-server feature "A detailed description of the feature to generate"

🔄 Environment Variables

Variable	Description	Default
`MCP_EXECUTION`	Required - Must be set to "true" for MCP mode	-
`OPENAI_API_KEY`	OpenAI API key for generating backlog	-
`GROQ_API_KEY`	Alternative Groq API key	-
`DEBUG`	Enable debug mode for additional logs	false
`TEST_MODE`	Enable test mode (mock generation)	false
`AGILE_PLANNER_OUTPUT_ROOT`	Base directory for output	current dir

📜 License

Agile Planner MCP Server is licensed under the MIT License with Commons Clause. See the LICENSE file for the complete license text.

👥 Support

For support, please open an issue on the GitHub repository or contact your Windsurf/Cascade/Cursor administrator.

☕️ Support the Project

If you find this project useful, you can support its development by buying me a coffee on BuyMeACoffee!

🚀 Get Windsurf

Thank you 🙏