Contributing to Microsoft's Azure DevOps MCP Server
Johan Spännare
When integrating AI agents with Azure DevOps for work item management, we identified a gap in the current MCP server implementation: the ability to update and delete work item comments was missing.
Rather than working around the limitation, we took a different approach.
The Problem
The Azure DevOps MCP server provides tools for listing and adding comments to work items. However, when AI agents manage work items autonomously, mistakes happen. Comments may contain errors, become outdated, or need to be removed entirely.
Without the ability to update or delete comments programmatically, users were forced to manually navigate to the Azure DevOps UI to make corrections.
Our Approach
Instead of waiting for someone else to address this, we:
- Documented the use case and submitted a feature request to the official repository
- Developed a complete implementation following the project’s existing patterns
- Published the solution as a standalone npm package for immediate use
The Implementation
The implementation adds two new tools:
wit_update_work_item_comment
- Update existing comments to correct errors or reflect progress
- Supports both markdown and HTML formatting
- Uses the Azure DevOps REST API PATCH endpoint
wit_delete_work_item_comment
- Permanently remove comments that are no longer relevant
- Uses the Azure DevOps REST API DELETE endpoint
Both tools follow the existing patterns in the codebase and include comprehensive tests.
Try It Now
While waiting for the feature request to be reviewed, the implementation is available as an npm package:
npm install -g @johanspannare/azure-devops-mcp-extended
To use it with Claude Code or other MCP-compatible tools, add this to your MCP configuration:
{
"mcpServers": {
"azure-devops": {
"command": "npx",
"args": ["-y", "@johanspannare/azure-devops-mcp-extended@latest", "<your-organization>"]
}
}
}
Links
Contributing Back
We believe in contributing back to the tools we rely on. Open source thrives when users become contributors. If you encounter a missing feature in a tool you depend on, consider implementing it yourself and sharing it with the community.