Skip to main content
Glama

az-devops-cli-mcp

Azure DevOps MCP tools for Claude Code and Codex CLI — boards, work items, repos, and PRs.

PyPI version CI Python 3.10+


Install & Register

# 1. Install
pip install az-devops-cli-mcp

# 2. Login (once per machine)
az login
az devops configure --defaults organization=https://dev.azure.com/YOUR_ORG

# 3. Register with Claude Code
claude mcp add az-devops-cli-mcp -- python -m az_devops_cli_mcp.server

# 4. Register with Codex CLI
codex mcp add az-devops-cli-mcp -- python -m az_devops_cli_mcp.server

Restart Claude Code or Codex CLI. All 13 tools are ready.


Related MCP server: @om-surushe/efficient-ado

How It Works

The server runs as a stdio process — spawned on demand by the client, no ports or background services needed.

Claude Code / Codex CLI
        |  spawns on demand
        v
python -m az_devops_cli_mcp.server   (stdin/stdout)
        |
        v
      az CLI  (your az login session)

Available Tools (13 total)

Work Items

Tool

What it does

create_work_item

Create Task, Bug, User Story, Epic

get_work_item

Get item by ID

update_work_item

Update state, assignment, title

set_iteration

Assign to sprint (required for Sprint board)

query_work_items

WIQL query for bulk lookups

link_work_items

Link items (Child, Parent, Related)

get_work_item_relations

Verify hierarchy after linking

add_comment

Add discussion comment / agent log

Repos / PRs

Tool

What it does

list_prs

List PRs by status

get_pr

Get PR details

create_pr

Open a new PR

link_pr_to_work_item

Trace PR to work item

list_branches

List repo branches


AG2 Usage (optional)

from autogen import ConversableAgent
from az_devops_cli_mcp import ALL_TOOLS, WORK_ITEM_TOOLS, REPO_TOOLS

agent = ConversableAgent(
    name="devops_agent",
    tools=ALL_TOOLS,
    system_message="You manage Azure DevOps for LytStore..."
)

Configuration

Method

How

az devops configure

Preferred — org set globally for your machine

AZURE_DEVOPS_ORG env var

Per-session override

.env file

AZURE_DEVOPS_ORG=https://dev.azure.com/YOUR_ORG


Safety

Destructive commands (delete, remove, destroy) are blocked at the core layer. The agent returns an error and requires explicit user confirmation.


Manual MCP Config (fallback)

If claude mcp add / codex mcp add aren't available, add this to your config file manually.

Claude Code — ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "az-devops-cli-mcp": {
      "type": "stdio",
      "command": "python",
      "args": ["-m", "az_devops_cli_mcp.server"]
    }
  }
}

Codex CLI — ~/.codex/config.json: same block.


Contributing

git clone https://github.com/PraiseSinkamba/az-devops-cli-mcp
cd az-devops-cli-mcp
pip install -e ".[dev]"
pytest tests/

Available Tools

13 tools
add_commentB

Add a discussion comment to a work item. Use to log agent reasoning or decisions.

Inputs: id (required), comment (required).

create_prC

Create a pull request.

Inputs: repo (required), title (required), source (required), project (required), +2 more.

create_work_itemB

Create a work item on the board. type options: Task, Bug, User Story, Epic, Feature.

Inputs: title (required), project (required), type, assigned_to.

get_prA

Get full details of a specific PR.

Inputs: pr_id (required).

get_work_itemC

Get a work item by ID.

Inputs: id (required).

get_work_item_relationsB

Get all linked work items. Use after bulk linking to verify hierarchy.

Inputs: id (required).

list_branchesC

List all branches in a repository.

Inputs: repo (required), project (required).

list_prsB

List pull requests in a project. status options: active, completed, abandoned, all.

Inputs: project (required), status.

query_work_itemsB

Run a WIQL query to find work items. Use for any bulk lookups.

Inputs: wiql (required), project (required).

set_iterationA

Assign a work item to a sprint iteration. Required for items to appear on the Sprint board. Format: 'ProjectName\Sprint N'

Inputs: id (required), iteration_path (required).

update_work_itemB

Update state, assignment, or title of a work item.

Inputs: id (required), state, title, assigned_to.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.4/5.0

    Scored across 13 tools

    Disambiguation5/5

    Each tool has a clearly defined purpose with no overlaps. Work item operations are distinct (create, get, update, query, link, set iteration, add comment), PR operations are separate, and branch listing is isolated.

    Naming Consistency5/5

    All tools follow a consistent verb_noun pattern in snake_case (e.g., create_work_item, list_prs, set_iteration). No mixing of conventions or inconsistent naming styles.

    Tool Count5/5

    13 tools is well-scoped for an Azure DevOps CLI, covering work items (create, read, update, query, link, set iteration, comment) and PRs (create, get, list, link). Each tool serves a clear purpose without excess.

    Completeness3/5

    Core operations are present (create/read/update for work items, create/read for PRs), but missing delete for work items, update/complete for PRs, and branch creation. Some gaps exist that may hinder full workflows.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers