@om-surushe/efficient-ado
Enables AI assistants to interact with Azure DevOps through an MCP server with 34 tools for PRs, work items, repos, and branches, optimized for token efficiency and agent-first workflows.
README
@om-surushe/efficient-ado
LLM-optimized Azure DevOps workflows for AI assistants
Efficient ADO is a Model Context Protocol (MCP) server that provides AI assistants with streamlined access to Azure DevOps. Built with agent-first design principles, it minimizes round trips, reduces token usage, and provides rich contextual responses.
Features
- 🎯 Agent-First Design - Composite tools reduce round trips
- âš¡ Token Efficient - Tiered responses (summary/standard/detailed)
- 🤖 LLM-Optimized - Rich context + suggested actions in every response
- 🚀 Fast - Built with TypeScript, optimized for performance
- 📦 Complete - 34 tools covering PRs, work items, repos, branches
- 🔄 Composite Workflows - Multi-step operations in single calls
- 🧠Error Guidance - Errors tell agents HOW to fix issues
Installation
npm install @om-surushe/efficient-ado
Quick Start
Configuration
Create a .env file:
AZDO_ORG_URL=https://dev.azure.com/your-org
AZDO_PAT=your-personal-access-token
AZDO_PROJECT=your-project # optional
AZDO_REPO=your-repo # optional
PAT Requirements:
- Code (Read & Write)
- Pull Request Threads (Read & Write)
- Work Items (Read & Write)
Usage with Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"efficient-ado": {
"command": "npx",
"args": ["@om-surushe/efficient-ado"],
"env": {
"AZDO_ORG_URL": "https://dev.azure.com/your-org",
"AZDO_PAT": "your-pat-here",
"AZDO_PROJECT": "your-project",
"AZDO_REPO": "your-repo"
}
}
}
}
Usage with OpenClaw / Cline / Other MCP Clients
Similar configuration - use stdio transport with environment variables.
Available Tools (34 Total)
Foundation
setup_workspace- Configure credentials and verify connection
Pull Request Workflows
Discovery:
list_prs- List PRs with filters (creator, reviewer, status)get_pr- Get PR details with reviewers and statusstart_review- COMPOSITE - Get PR + comments + reviewers in one call
Review & Approval:
vote_on_pr- Approve (10), approve with suggestions (5), wait (-5), reject (-10)manage_reviewers- Add/remove reviewersquick_approve- COMPOSITE - Approve + merge if ready
Merge:
check_merge_readiness- Verify all merge requirementscomplete_pr- Merge PR with completion options
Comments & Threads:
list_comments- List all threads (general + inline)add_comment- Add general or inline commentreply_to_thread- Reply to existing threadupdate_thread_status- Mark thread as fixed/active/wontFix/closedaddress_comment- COMPOSITE - Reply + resolve in one call
Files & Diffs:
get_pr_changes- List changed files with statsget_file_diff- Get file diff (side-by-side comparison)get_file_content- Get file content at PR commit
PR Lifecycle:
create_pr- Create PR with work items, reviewers, auto-completeupdate_pr- Update title, description, draft statusabandon_pr- Close PR without mergingreactivate_pr- Reopen abandoned PR
Work Items
get_work_item- Get work item by ID with relationslist_my_work_items- List work items assigned to youcreate_work_item- Create new task/bug/storyupdate_work_item- Update work item fieldsadd_work_item_comment- Add discussion commentlink_work_item- Link work items (parent/child/related)
Repos & Branches
list_repositories- List repos in projectlist_branches- List branches in repocreate_branch- Create new branch from sourcelist_commits- List commits with filtersget_commit- Get commit detailscompare_branches- Compare two branches
Daily Workflow
get_my_work- COMPOSITE - Dashboard: my PRs + reviews + work items
Usage Examples
Daily Standup
> "What do I need to work on today?"
Agent uses: get_my_work()
Response includes:
- PRs you created (3 active, 1 draft)
- PRs needing your review (2 pending vote)
- Work items assigned to you (5 tasks, 2 bugs)
Code Review
> "Review PR #1805"
Agent uses: start_review(prId=1805)
Gets in one call:
- PR details (title, description, status, branches)
- All comment threads (23 general + 5 inline)
- Reviewers and their votes
- Changed files overview
Then:
> "Show me the changes to config.ts"
Agent uses: get_file_diff(prId=1805, path="config.ts")
Returns side-by-side diff with line numbers
Addressing Feedback
> "Reply 'Fixed indentation' and mark thread 12345 as resolved"
Agent uses: address_comment(
prId=1805,
threadId=12345,
reply="Fixed indentation",
status="fixed"
)
One call handles both reply and status update
Approving & Merging
> "Approve PR #1805 and merge if ready"
Agent uses: quick_approve(prId=1805)
One call:
1. Approves PR (vote=10)
2. Checks merge requirements
3. Merges if all policies pass
4. Or tells you what's blocking
Creating Work from Feedback
> "Create a bug for the memory leak in thread 12345"
Agent uses:
1. list_comments(prId=1805) to get thread details
2. create_work_item(
type="Bug",
title="Memory leak in cache handler",
description="Details from PR review..."
)
3. link_work_item(workItemId=123, prId=1805)
Three calls link bug to the PR discussion
Design Philosophy
Agent-First
Every tool response includes:
- Current state - What's happening now
- Statistics - Quick metrics
- Blockers - What's preventing progress (if any)
- Suggested actions - What to do next with exact tool names
Token Efficient
Responses have levels:
- Summary (~200 tokens) - Quick status check
- Standard (~800 tokens) - Most common use
- Detailed (~3000 tokens) - When agent needs everything
Error Messages That Help
Errors tell agents HOW to fix issues:
{
"error": {
"code": "MERGE_BLOCKED",
"blockers": [
{
"reason": "Insufficient approvals (1/2)",
"howToFix": "Call vote_on_pr(vote=10) to approve",
"relatedTools": ["vote_on_pr"]
}
]
}
}
Development
git clone https://github.com/om-surushe/efficient-ado.git
cd efficient-ado
npm install
npm run build
Status
- [x] Phase 1: Foundation + setup
- [x] Phase 2: Pull request workflows (15 tools)
- [x] Phase 3: Work items (6 tools)
- [x] Phase 4: Repos & branches (6 tools)
- [x] Phase 5: PR creation & update (4 tools)
- [x] Phase 6: Composite workflows (2 tools)
Total: 34 tools complete and ready for use! 🎉
Future Enhancements
- [ ] Pipeline/build tools
- [ ] Team/project management
- [ ] Wiki/documentation tools
- [ ] Artifact management
- [ ] Test plan integration
License
MIT License - see LICENSE for details.
Author
Om Surushe
- GitHub: @om-surushe
- npm: @om-surushe
Related Packages
- @om-surushe/efficient-ticktick - LLM-optimized TickTick task management
- @om-surushe/efficient-search - LLM-optimized web search
Part of the Efficient MCP series - LLM-optimized tools for AI assistants.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.