Agile MCP
A comprehensive agile backlog management system built with the Model Context Protocol (MCP), featuring project isolation, multi-agent support, and interactive visualization
Overview
Agile MCP provides a complete agile backlog management solution that bridges AI agents and human teams. It consists of three integrated components:
MCP Server: Exposes agile backlog tools via Model Context Protocol for AI agents (Claude, etc.)
Web UI: Interactive dashboard for humans with dependency graphs and hierarchy trees
Shared Database: SQLite database with project isolation and security auditing
Key Features
š Project Isolation
Identifier-based security: Each project has a unique identifier
Multi-project support: Manage multiple repositories in a single database
Access control: AI agents can only access items from their registered project
Security auditing: All access attempts logged for compliance
š¤ MCP Server (AI Agent Interface)
Epic, Story, and Task Management: Full CRUD operations
Sprint & Iteration Planning: Create sprints, manage capacity, track velocity
Burndown Tracking: Daily snapshots and burndown chart data
Dependency Tracking: Create dependencies with circular dependency prevention
Conflict Detection: Track modifications to detect concurrent edits
Export Capabilities: Export backlog data for analysis
MCP Resources: Read-only views for backlog overview and dependencies
š Web UI (Human Interface)
Project Management: Register and switch between projects
Full CRUD Operations: Create, edit, and delete epics, stories, and tasks
Sprint Management: Create and manage sprints, add/remove stories
Sprint Board: Kanban-style board with drag-and-drop (5 status columns)
Burndown Charts: Visual burndown tracking with ideal vs actual lines
Velocity Reports: Historical velocity calculation for sprint planning
Backlog List View: Filterable list with sprint and epic filtering
Dependency Graph: Interactive React Flow visualization showing story dependencies
Hierarchy Tree: D3-powered tree view of Epic ā Story ā Task relationships
Responsive Design: Mobile-friendly interface with adaptive layouts
š¾ Database & Security
SQLite with WAL mode: Fast, reliable local storage
Foreign key constraints: Cascade deletes maintain data integrity
Automatic migrations: Database schema updates applied automatically
Security logs: Audit trail for unauthorized access attempts
Conflict detection: Track concurrent modifications
Architecture
Quick Start
Prerequisites
Node.js 18+
npm 9+
Installation
Usage
For AI Agents (MCP Server)
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
Then in Claude:
For Humans (Web UI)
Visit http://localhost:3004 to access:
Backlog List (
/) - Filterable story list with CRUD operations and sprint filteringSprint Board (
/project/:id/sprint/:sprintId) - Kanban board for active sprintsBurndown Chart (
/project/:id/sprint/:sprintId/burndown) - Sprint burndown visualizationDependency Graph (
/project/:id/dag) - Interactive dependency visualizationHierarchy Tree (
/project/:id/tree) - Epic ā Story ā Task hierarchy
Core Concepts
Projects
Each project represents a codebase/repository with its own isolated backlog:
Unique identifier (e.g.,
"frontend-app","api-service")All epics, stories, and tasks belong to a project
AI agents must register projects before creating backlog items
Epics, Stories, and Tasks
Standard agile hierarchy:
Epics: High-level feature areas or initiatives
Stories: User stories with priority, points, and status
Tasks: Actionable sub-items of stories
Sprints & Iterations
Time-boxed work periods with capacity planning:
Sprint lifecycle: planning ā active ā completed
Story assignment: Add/remove stories from sprints
Capacity tracking: Monitor committed vs completed points
Burndown charts: Track remaining work over time
Velocity metrics: Calculate historical team velocity
Daily snapshots: Record progress for burndown visualization
Dependencies
Story-to-story relationships:
blocks / blocked_by dependency types
Automatic circular dependency detection
Visualized in dependency graph view
Status Workflow
All items follow this workflow:
Technology Stack
Layer | Technology |
MCP Protocol | @modelcontextprotocol/sdk |
Database | SQLite (better-sqlite3) |
Backend API | Express.js, TypeScript |
Frontend | React 18, TypeScript, Vite |
Graph Visualization | React Flow (@xyflow/react) |
Tree Visualization | D3.js |
Styling | Tailwind CSS |
Icons | Lucide React |
Documentation
QUICKSTART.md - Get started in 5 minutes
ARCHITECTURE.md - Technical architecture and design
MCP_SERVER.md - Complete MCP tool reference for AI agents
API_REFERENCE.md - REST API documentation for developers
DEPLOYMENT.md - Production deployment guide
Example Workflows
AI Agent Workflow
Web UI Workflow
Select project from dropdown (or create new project)
Click "New Epic" to create an epic
Click "New Story" to create a story
Click "New Sprint" to create a sprint for iteration planning
Add stories to sprints from story detail pages
View active sprint board by clicking "Active Sprint" button
Track progress with burndown charts
Use edit/delete icons for modifications
Switch to "Dependency Graph" to visualize relationships
Switch to "Hierarchy Tree" to see the full hierarchy
Security Features
Project Isolation
Each project has a unique identifier
All MCP tools validate project access
Cross-project access attempts are blocked and logged
Audit Trail
Security logs track all access violations
Event types:
unauthorized_access,project_violation,conflict_detectedAccessible via
/api/security-logsendpoint
Conflict Detection
Tracks
last_modified_byfor all entitiesDetects concurrent modifications
Warns when multiple agents edit the same item
Development
Build All Packages
Watch Mode
Database Location
The SQLite database is created at ./agile-backlog.db in the project root.
Contributing
Contributions are welcome! Please:
Fork the repository
Create a feature branch
Make your changes
Submit a pull request
License
MIT License - see LICENSE file for details
Support
Issues: Report bugs or request features via GitHub issues
Documentation: See docs/ directory for detailed guides
MCP SDK: https://github.com/anthropics/mcp
Built with ā¤ļø for seamless AI-human collaboration in agile development
This server cannot be installed
local-only server
The server can only run on the client's local machine because it depends on local resources.
Enables comprehensive agile project management with epics, stories, tasks, and sprint planning through natural language. Features project isolation, dependency tracking, and integrates with a web UI for visual backlog management and burndown charts.