Azure DevOps MCP Server

This MCP (Model Context Protocol) server provides tools for interacting with Azure DevOps services through AI assistants.

Architecture

The server follows an entity-based architecture that groups operations by resource type rather than exposing many atomic tools. This approach provides several benefits:

1. Intuitive organization: Tools are organized by the entities they operate on (projects, repositories, work items, etc.)
2. Reduced tool count: Instead of dozens of individual tools, we have a handful of entity tools with multiple operations
3. Consistent interface: All entity tools follow the same pattern for operations and parameters
4. Better error handling: Each entity tool can handle errors specific to its domain
5. Easier discovery: Users can easily discover available operations for each entity

Architecture Diagram

flowchart TB
    Client[AI Assistant] -->|MCP Request| Server[MCP Server]
    Server -->|MCP Response| Client
    
    subgraph "Azure DevOps MCP Server"
        Server --> RequestHandler[Request Handler]
        RequestHandler --> ToolRegistry[Tool Registry]
        ToolRegistry --> EntityTools[Entity Tools]
        EntityTools --> ApiClient[API Client]
        ApiClient --> ErrorUtils[Error Utilities]
        ApiClient --> PaginationUtils[Pagination Utilities]
        ApiClient -->|HTTP Request| AzureDevOps[Azure DevOps API]
        AzureDevOps -->|HTTP Response| ApiClient
        ConfigManager[Configuration Manager] --> ApiClient
    end
    
    classDef primary fill:#4285F4,stroke:#0D47A1,color:white
    classDef secondary fill:#34A853,stroke:#0D652D,color:white
    classDef utility fill:#FBBC05,stroke:#866A00,color:white
    classDef external fill:#EA4335,stroke:#980905,color:white
    
    class Server,RequestHandler primary
    class ToolRegistry,EntityTools,ApiClient secondary
    class ErrorUtils,PaginationUtils,ConfigManager utility
    class Client,AzureDevOps external

Component Structure

classDiagram
    class EntityTool {
        +name: string
        +description: string
        +operations: Record~string, Function~
        +schemas: Record~string, ZodSchema~
        +getDefinition(): ToolDefinition
        +execute(args: unknown): Promise~any~
        #registerOperation(operation, handler, schema, description)
    }
    
    class ADOApiClient {
        +config: ADOApiConfig
        +connection: WebApi
        +getCoreApi()
        +getWorkItemTrackingApi()
        +getGitApi()
        +getPipelineApi()
        +handleError(error, context)
    }
    
    class ToolRegistry {
        +registerTool(tool: Tool)
        +getTool(name: string): Tool
        +getToolDefinitions(): ToolDefinition[]
    }
    
    class ErrorUtils {
        +createError(code, message, context)
        +handleApiError(error, source, operation)
    }
    
    class PaginationUtils {
        +normalizePaginationParams(params)
        +createPaginationResult(items, totalCount, continuationToken)
        +encodeContinuationToken(data)
        +decodeContinuationToken(token)
    }
    
    EntityTool --> ADOApiClient : uses
    EntityTool --> ErrorUtils : uses
    EntityTool --> PaginationUtils : uses
    ToolRegistry --> EntityTool : registers

Key Components

• Entity Tools: Each tool represents a major Azure DevOps entity (projects, repositories, work items, etc.) and provides multiple operations (list, get, create, etc.)
• Tool Registry: Manages the registration and execution of entity tools
• API Client: Handles communication with the Azure DevOps REST API
• Error Utilities: Provides standardized error handling with detailed context and user-friendly messages
• Pagination Utilities: Implements cursor-based pagination for list operations
• Configuration Manager: Loads and validates configuration from environment variables or config file

Recent Improvements

1. Enhanced Error Handling

The server now includes a comprehensive error handling system that provides:

• Categorized errors: Errors are categorized by type (authentication, authorization, validation, etc.)
• Contextual information: Errors include the source, operation, and other relevant context
• User-friendly messages: Error messages are designed to be helpful and actionable
• Troubleshooting tips: Where applicable, errors include suggestions for resolving the issue

flowchart LR
    Error[API Error] --> Handler[Error Handler]
    Handler --> Category{Categorize}
    Category -->|Authentication| AuthError[Authentication Error]
    Category -->|Authorization| AuthzError[Authorization Error]
    Category -->|Not Found| NotFoundError[Not Found Error]
    Category -->|Validation| ValidationError[Validation Error]
    Category -->|Rate Limit| RateLimitError[Rate Limit Error]
    Category -->|Service| ServiceError[Service Error]
    Category -->|Unknown| UnknownError[Unknown Error]
    
    AuthError & AuthzError & NotFoundError & ValidationError & RateLimitError & ServiceError & UnknownError --> Format[Format User Message]
    Format --> McpError[MCP Error Response]
    
    classDef error fill:#EA4335,stroke:#980905,color:white
    classDef process fill:#4285F4,stroke:#0D47A1,color:white
    classDef result fill:#34A853,stroke:#0D652D,color:white
    
    class Error,AuthError,AuthzError,NotFoundError,ValidationError,RateLimitError,ServiceError,UnknownError error
    class Handler,Category,Format process
    class McpError result

2. Cursor-based Pagination

All list operations now support cursor-based pagination with:

• Continuation tokens: Encoded tokens for resuming pagination
• Customizable page size: Control the number of results per page
• Consistent interface: Same pagination parameters across all list operations
• Efficient resource usage: Only fetch the data you need

sequenceDiagram
    participant Client as AI Assistant
    participant Server as MCP Server
    participant API as Azure DevOps API
    
    Client->>Server: List request (maxResults=10)
    Server->>API: API request (top=10, skip=0)
    API->>Server: Response with items
    
    Note over Server: Create continuation token
    
    Server->>Client: Response with items and token
    
    Client->>Server: List request with token
    
    Note over Server: Decode token to get position
    
    Server->>API: API request (top=10, skip=10)
    API->>Server: Response with more items
    Server->>Client: Response with items and new token

3. Improved Documentation

Each tool and operation now includes:

• Detailed descriptions: Clear explanations of what each tool and operation does
• Parameter documentation: Comprehensive documentation for all parameters
• Usage examples: Real-world examples of how to use each operation
• Type information: Clear type definitions for all inputs and outputs

Available Entity Tools

Projects Tool

Manages Azure DevOps projects.

Operations:

• list: List all projects in the organization with pagination support
• get: Get detailed information about a specific project

Repositories Tool

Manages Git repositories.

Operations:

• list: List all Git repositories in a project with pagination support
• get: Get detailed information about a specific Git repository
• listBranches: List all branches in a Git repository with pagination support

Work Items Tool

Manages work items (bugs, tasks, user stories, etc.).

Operations:

• get: Get detailed information about a specific work item
• create: Create a new work item in a project

Pull Requests Tool

Manages pull requests in repositories.

Operations:

• list: List pull requests in a repository with filtering and pagination support
• get: Get detailed information about a specific pull request

Pipelines Tool

Manages CI/CD pipelines.

Operations:

• list: List all pipelines in a project with pagination support
• get: Get detailed information about a specific pipeline

Usage Examples

List Projects with Pagination

{
  "operation": "list",
  "listParams": {
    "maxResults": 10,
    "continuationToken": "optional-token-from-previous-request"
  }
}

Get Project Details

{
  "operation": "get",
  "getParams": {
    "projectId": "my-project",
    "includeCapabilities": true
  }
}

List Repositories in a Project

{
  "operation": "list",
  "listParams": {
    "projectId": "my-project",
    "maxResults": 20
  }
}

List Branches in a Repository

{
  "operation": "listBranches",
  "listBranchesParams": {
    "projectId": "my-project",
    "repositoryId": "my-repo",
    "maxResults": 15
  }
}

Get Work Item Details

{
  "operation": "get",
  "getParams": {
    "id": 123,
    "expand": "Relations"
  }
}

Create a Work Item

{
  "operation": "create",
  "createParams": {
    "projectId": "my-project",
    "type": "Task",
    "title": "Implement new feature",
    "description": "This task involves implementing the new feature XYZ",
    "assignedTo": "[email protected]"
  }
}

List Pull Requests with Filtering

{
  "operation": "list",
  "listParams": {
    "projectId": "my-project",
    "repositoryId": "my-repo",
    "status": "Active",
    "maxResults": 10
  }
}

Configuration

The server can be configured using environment variables or a configuration file.

Environment Variables

• ADO_ORGANIZATION: Azure DevOps organization name (required)
• ADO_PROJECT: Default project name (optional)
• ADO_PAT: Personal Access Token for authentication (required)
• ADO_API_URL: Base URL for the API (optional, defaults to https://dev.azure.com)
• ADO_API_VERSION: API version (optional, defaults to 7.0)
• ADO_API_MAX_RETRIES: Maximum number of retries for API calls (optional, defaults to 3)
• ADO_API_DELAY_MS: Delay between retries in milliseconds (optional, defaults to 1000)
• ADO_API_BACKOFF_FACTOR: Backoff factor for retries (optional, defaults to 2)

Configuration File

Alternatively, you can create a config/azuredevops.json file with the following structure:

{
  "organization": "your-organization",
  "project": "your-project",
  "credentials": {
    "pat": "your-personal-access-token"
  },
  "api": {
    "baseUrl": "https://dev.azure.com",
    "version": "7.0",
    "retry": {
      "maxRetries": 3,
      "delayMs": 1000,
      "backoffFactor": 2
    }
  }
}

Development

Building the Server

npm run build

Running the Server

node build/index.js

Docker

docker build -t azure-devops-mcp:local .
docker run -i --rm -e ADO_ORGANIZATION=your-org -e ADO_PAT=your-pat azure-devops-mcp:local

## License

MIT License © 2025 Aaron Bockelie <[email protected]>