triagent 1.4.0rc22

Claude-powered CLI for Azure DevOps automation

Triagent CLI

Claude-powered CLI for Azure DevOps automation.

Overview

Triagent is an interactive command-line tool that uses Claude AI (via Azure AI Foundry) to help automate Azure DevOps operations.

Capabilities

• Azure DevOps Automation

  • Create, update, and query work items
  • Create, review, and manage Pull Requests
  • Monitor build and release pipelines
  • Add PR comments and set votes

• Kusto Log Analysis

  • Generate Kusto queries for Application Insights
  • Query AppExceptions, AppRequests, AppDependencies, AppTraces
  • Support for multiple Log Analytics workspaces (DEV/QAS/PRD)

• Defect Investigation

  • Automatic investigation workflow for defects/incidents
  • Service-to-AppRoleName mapping
  • Telemetry correlation across tables

• Security Controls

  • Pre-tool execution validation (blocks dangerous commands)
  • Write operation confirmations (configurable)
  • Azure CLI and Git operation guards

Architecture

Triagent uses the Claude Agent SDK to provide an AI-powered interactive CLI:

┌─────────────────────────────────────────────────────────┐
│                    Triagent CLI                         │
├─────────────────────────────────────────────────────────┤
│  Interactive REPL │ Slash Commands │ Activity Tracker   │
├─────────────────────────────────────────────────────────┤
│                 Claude Agent SDK                        │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────────┐  │
│  │ SDK Client  │  │  Security   │  │  System Prompt  │  │
│  │   Options   │  │   Hooks     │  │   + Team CLAUDE │  │
│  └─────────────┘  └─────────────┘  └─────────────────┘  │
├─────────────────────────────────────────────────────────┤
│                   MCP Servers                           │
│  ┌─────────────────────┐  ┌─────────────────────────┐   │
│  │  triagent (local)   │  │  azure-devops (npx)     │   │
│  │  - get_team_config  │  │  - work items           │   │
│  │  - generate_kusto   │  │  - pull requests        │   │
│  │  - list_tables      │  │  - pipelines            │   │
│  └─────────────────────┘  └─────────────────────────┘   │
└─────────────────────────────────────────────────────────┘

Components

• TriagentSDKClient: Builds ClaudeAgentOptions with hooks, MCP servers, and team-specific prompts
• Security Hooks: Pre/post-tool execution validation (blocks dangerous commands, confirms writes)
• MCP Tools: In-process triagent tools + external Azure DevOps MCP server
• Team Configurations: Custom CLAUDE.md files per team with ADO context

Installation

Quick Install (Recommended)

macOS/Linux:

curl -sSL https://raw.githubusercontent.com/sdandey/triagent/main/install.sh | bash

Windows (PowerShell):

irm https://raw.githubusercontent.com/sdandey/triagent/main/install.ps1 | iex

Alternative Methods

# Using pipx (recommended for CLI tools)
pipx install triagent

# Using uv
uv tool install triagent

# Using pip
pip install triagent

Install Specific Version

# macOS/Linux
curl -sSL https://raw.githubusercontent.com/sdandey/triagent/main/install.sh | bash -s -- --version 0.2.0

# Windows
.\install.ps1 -Version "0.2.0"

Quick Start

# Start interactive chat
triagent

# Run setup wizard
triagent
> /init

Web Terminal (Docker)

Run triagent in a browser-based terminal - no local Python installation required.

One-Liner Install (Recommended)

Windows (PowerShell):

irm https://raw.githubusercontent.com/sdandey/triagent/main/start-web.ps1 | iex

macOS/Linux:

curl -sSL https://raw.githubusercontent.com/sdandey/triagent/main/start-web.sh | bash

This downloads, starts the container, and opens http://localhost:7681 in your browser.

Manual Setup

# Build and start (from cloned repo)
docker compose -f docker-compose.web.yml up -d

# Open in browser
open http://localhost:7681

# Stop
docker compose -f docker-compose.web.yml down

See docs/web-terminal.md for more details.

Setup Wizard

The /init command guides you through configuration:

1. Model Selection - Choose API provider (Azure Foundry, Anthropic)
2. Team Selection - Choose team and ADO project
3. MCP Server Setup - Configure Azure DevOps MCP server
4. Azure Authentication - Browser-based Azure login
5. Prerequisites Check - Verify required tools are installed

Note: The wizard displays installation instructions but does not auto-install tools. See Prerequisites for installation instructions.

Slash Commands

Command	Description
/init	Run setup wizard
/help	Show available commands
/config	View current configuration
/config set <key> <value>	Set a config value
/team	Show current team
/team <name>	Switch team
/clear	Clear conversation history
/exit	Exit Triagent

MCP Tools

Triagent provides custom MCP tools for team-specific operations:

Tool	Description
get_team_config	Returns team configuration for Kusto queries and ADO context
generate_kusto_query	Generates Kusto query templates for Application Insights
list_telemetry_tables	Lists available telemetry tables with field descriptions

External MCP server for Azure DevOps operations:

• @anthropic-ai/mcp-server-azure-devops@0.1.1

Security

Triagent implements security hooks to protect against dangerous operations:

Blocked Commands

• Destructive bash patterns (rm -rf /, DROP TABLE, fork bombs)
• Direct file system attacks (> /dev/, chmod 777)

Confirmed Operations

• Azure DevOps writes (work items, PRs, comments)
• Azure CLI writes (pipelines, repos, policies)
• Git operations (commit, push, merge)

Toggle confirmations: /confirm off (use with caution)

Configuration

Configuration is stored in ~/.triagent/:

~/.triagent/
├── config.json          # Main configuration
├── credentials.json     # API credentials (secure)
├── mcp_servers.json     # MCP server config
└── history/             # Conversation history

Session Logs

Triagent logs session activity for troubleshooting and debugging. Logs are stored in platform-specific locations:

Platform	Location
macOS	~/Library/Logs/triagent/sessions/
Windows	%LOCALAPPDATA%\triagent\logs\sessions\
Linux	~/.local/share/triagent/logs/sessions/

Each session creates a timestamped log file (e.g., session_20260103_143022.log) containing:

• Session start/end events
• User prompts
• Tool calls (MCP tools, bash commands)
• Agent responses and errors

Example Windows path: C:\Users\{username}\AppData\Local\triagent\logs\sessions\

Teams

Triagent supports multiple team configurations with custom CLAUDE.md files:

Team	Display Name	ADO Project	Features
levvia	Levvia	Project Omnia	Custom prompts
omnia	Omnia	Project Omnia	Custom prompts
omnia-data	Omnia Data	Audit Cortex 2	18 repos, service mappings, telemetry config

Switch teams: /team <name>

Development

# Clone repository
git clone https://github.com/sdandey/triagent.git
cd triagent

# Install with UV
uv pip install -e ".[dev]"

# Run tests
pytest tests/ -v

# Lint
ruff check src/

Prerequisites

Required

• Python 3.11+ - Core runtime
• Node.js 18+ - For MCP servers (Azure DevOps integration)
• Git for Windows - (Windows only) Required by the bundled Claude Code CLI
• Azure DevOps account - For ADO automation

Note: Claude Code CLI is automatically bundled with the claude-agent-sdk package. No separate npm installation is required.

Recommended

• Azure CLI - For Azure authentication and ADO operations

  • macOS: brew install azure-cli
  • Windows: Download from https://aka.ms/installazurecliwindows
  • Linux: curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash

• Azure CLI Extensions - Install after Azure CLI:

  az extension add --name azure-devops
  az extension add --name application-insights --allow-preview true
  az extension add --name log-analytics --allow-preview true
  

API Access (choose one)

• Azure AI Foundry
• Direct Anthropic API

Platform Setup

macOS

1. Install Homebrew (if not installed):

   /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"
   

2. Install Python 3.11+:

   brew install python@3.11
   

3. Install Node.js 18+:

   brew install node
   

4. Install Azure CLI:

   brew install azure-cli
   

5. Install Triagent:

   pip install triagent
   # or use the installer script:
   curl -sSL https://raw.githubusercontent.com/sdandey/triagent/main/install.sh | bash
   

6. Configure Azure CLI extensions:

   az extension add --name azure-devops
   az extension add --name application-insights --allow-preview true
   az extension add --name log-analytics --allow-preview true
   

7. Login to Azure:

   az login
   

Windows

1. Install prerequisites via PowerShell (Run as Administrator):

   irm https://raw.githubusercontent.com/sdandey/triagent/main/install.ps1 | iex
   

   This installs Python, Node.js, Git, and Azure CLI.

2. Install triagent:

   pip install triagent
   

3. Configure Azure CLI extensions (PowerShell):

   az extension add --name azure-devops
   az extension add --name application-insights --allow-preview true
   az extension add --name log-analytics --allow-preview true
   

4. Login to Azure:

   az login
   

5. Run triagent:

   triagent
   

Git Bash Users: If you want to run triagent from Git Bash, you need to set environment variables in ~/.bashrc. See docs/windows-installation.md for detailed instructions.

Manual Installation (Alternative)

If you prefer manual installation:

1. Python 3.11+: Download from https://www.python.org/downloads/

   • Check "Add Python to PATH" during installation

2. Node.js 18+: Download from https://nodejs.org/

   • Use LTS version

3. Git for Windows: Download from https://git-scm.com/download/win

   • Required by the bundled Claude Code CLI

4. Azure CLI (optional): Download from https://aka.ms/installazurecliwindows

5. Install triagent: pip install triagent

Verify Installation

Run these commands to verify your setup:

# Check versions
triagent --version
python --version
node --version
az --version  # Optional

# Start triagent and run setup wizard
triagent
> /init

License

MIT License - see LICENSE for details.