epam/dm.ai/dmtools-ai-docs

DMtools Development Assistant

DMTools is an enterprise dark-factory orchestrator that integrates with multiple platforms and provides 413+ MCP tools for reusable delivery automation.

🔧 FIRST-TIME SETUP (DO THIS PROACTIVELY)

When a user mentions DMtools or asks to use it, IMMEDIATELY perform these checks:

🚀 Quick Installation Bootstrap (Recommended)

Install or refresh DMtools from the latest release first, then continue with the configuration checks below:

# Install or update DMtools from the latest release
curl -fsSL https://github.com/epam/dm.ai/releases/latest/download/install.sh | bash

Then make sure to:

1. ✓ Check whether dmtools is available
2. ✓ Create dmtools.env if missing
3. ✓ Add DMtools files to .gitignore
4. ✓ Verify Java installation
5. ✓ Test DMtools functionality with dmtools list

OR follow manual steps below:

Focused skill packages: install only the integrations you need with DMTOOLS_SKILLS=jira,github and the skill installer described in references/installation/README.md#install-only-the-skills-you-need.

Child agent configs: define a shared base config and extend it with the existing `parent` inheritance pattern.

Step 1: Check if DMtools is installed

# Check if dmtools command is available
which dmtools || echo "DMtools not installed"

If NOT installed:

# Offer to install DMtools automatically
curl -fsSL https://github.com/epam/dm.ai/releases/latest/download/install.sh | bash

Step 2: Check for dmtools.env configuration

# Check if dmtools.env exists in current directory or home
ls dmtools.env ~/.dmtools.env 2>/dev/null || echo "No dmtools.env found"

If NOT found, help user create it:

1. Ask user which integrations they need (Jira, ADO, Figma, etc.)
2. Ask user which AI provider they want (Gemini free tier recommended)
3. Create dmtools.env with appropriate template (see Configuration Template below)
4. Add dmtools.env to .gitignore (contains secrets!)

Step 3: Verify .gitignore

# Check if dmtools files are in .gitignore
grep -q "dmtools.env\|dmtools-local.env" .gitignore 2>/dev/null || echo "Not in .gitignore"

If NOT in .gitignore, add these lines:

# DMtools configuration (contains secrets)
dmtools.env
dmtools-local.env

Configuration Template

When creating dmtools.env, use this template and fill with user's credentials:

# dmtools.env - NEVER commit this file
# Generated by DMtools skill assistant
 
# ================================
# INTEGRATIONS (choose what you need)
# ================================
 
# Jira (if user needs Jira)
JIRA_BASE_PATH=https://your-company.atlassian.net
JIRA_EMAIL=your-email@company.com
JIRA_API_TOKEN=your-jira-api-token
JIRA_AUTH_TYPE=Basic
 
# Azure DevOps (if user needs ADO)
# ADO_BASE_PATH=https://dev.azure.com/your-org
# ADO_PAT_TOKEN=your-ado-pat-token
 
# ================================
# AI PROVIDER (required - choose one)
# ================================
 
# Gemini (RECOMMENDED - free tier available)
GEMINI_API_KEY=your-gemini-api-key
# Get free key: https://aistudio.google.com/app/apikey
 
# OR OpenAI
# OPENAI_API_KEY=your-openai-api-key
 
# OR AWS Bedrock
# BEDROCK_ACCESS_KEY_ID=your-aws-key
# BEDROCK_SECRET_ACCESS_KEY=your-aws-secret
 
# ================================
# DEFAULTS
# ================================
DEFAULT_LLM=gemini
DEFAULT_TRACKER=jira

Quick Links for API Tokens:

• Jira: https://id.atlassian.com/manage-profile/security/api-tokens
• Gemini: https://aistudio.google.com/app/apikey (FREE - 15 req/min)
• OpenAI: https://platform.openai.com/api-keys
• ADO: https://dev.azure.com → User Settings → Personal Access Tokens

Step 4: Test Installation

# Verify installation works
dmtools list | head -5

When to Use

Use this skill when:

• Installing or configuring DMtools (do setup steps above first!)
• Setting up integrations (Jira, Azure DevOps, Figma, Confluence, Teams)
• Configuring AI providers (Gemini, OpenAI, Claude, DIAL, Ollama)
• Developing JavaScript agents with MCP tools
• Generating test cases (Jira, Xray, Cucumber)
• Generating analytics reports (ReportGenerator, ReportVisualizer)
• Troubleshooting DMtools issues
• Working with dmtools.env configuration
• Creating AI teammate configurations (Teammate/CliAgent)
• Setting up CI/CD run tracing (ciRunUrl) for ticket traceability

Quick Reference

⚠️ IMPORTANT: If this is first time using DMtools, see FIRST-TIME SETUP section above.

Manual Installation Steps

If automated setup didn't work, follow these manual steps:

1. Install DMtools CLI

curl -fsSL https://github.com/epam/dm.ai/releases/latest/download/install.sh | bash

2. Create dmtools.env

Create dmtools.env in project root (see template in FIRST-TIME SETUP section)

3. Add to .gitignore

echo -e "\n# DMtools configuration\ndmtools.env\ndmtools-local.env" >> .gitignore

4. Get API Tokens

• Jira: https://id.atlassian.com/manage-profile/security/api-tokens
• Gemini (FREE): https://aistudio.google.com/app/apikey
• OpenAI: https://platform.openai.com/api-keys

See Installation Guide for detailed setup.

Common Commands

dmtools list                          # List all 413+ MCP tools
dmtools jira_get_ticket PROJ-123      # Get Jira ticket
dmtools run agents/config.json        # Run configuration
dmtools run agents/config.json --ciRunUrl "https://ci.example.com/runs/42"  # With CI tracing
dmtools run agents/config.json "${ENCODED_CONFIG}" --inputJql "key=PROJ-1"  # With overrides

Core Capabilities

413+ MCP Tools Available

Complete Reference: references/mcp-tools/README.md - Auto-generated from actual DMtools build

Current breakdown (24 integrations):

• Jira (60 tools): Ticket management, search, comments, fields
• Jira Xray (11 tools): Xray test management
• Teams (28 tools): Messages, chats, files, transcripts, meetings
• Teams Auth (3 tools): Teams authentication flows
• Confluence (24 tools): Page management, search, content access, attachments, inline comments
• Rovo (4 tools): Atlassian Rovo MCP semantic search, Confluence AI summaries, Teamwork Graph context, and Q&A
• ADO (45 tools): Azure DevOps work items, queries, comments, attachments, pull requests, code review threads
• GitHub (77 tools): Pull requests, issues, comments, workflows
• GitLab (34 tools): Merge requests, issues, CI/CD pipelines
• Bitbucket (1 tool): Pull request operations
• Figma (22 tools): Design extraction, icons, layers, styles, components
• AI Providers (15 tools): Chat, vision, and file inputs across Gemini, OpenAI, Anthropic, Bedrock, DIAL, Ollama, and Vertex
• Knowledge Base (5 tools): Document search, indexing, RAG
• File (6 tools): File operations, read/write/list/exists
• Mermaid (3 tools): Diagram generation
• SharePoint (2 tools): Document management
• Bitrise (24 tools): Mobile CI/CD builds and artifacts
• Jenkins (7 tools): CI/CD job and build information
• TestRail (16 tools): Test case and test run management
• Rally (1 tool): Work item lookups
• SCM (15 tools): Vendor-neutral pull/merge request, label, comment and issue aliases routed by DEFAULT_SCM (github|gitlab)
• CI (4 tools): Vendor-neutral run trigger/list, verdict and merge-state aliases routed by DEFAULT_CI (actions|gitlab-ci)
• ENV (2 tools): Read-only access to non-secret configuration values for JS agents (env_get, env_list); credentials are never returned, extend the list with JS_ENV_ALLOWLIST
• CLI (1 tool): Command execution

Example tools:

• jira_get_ticket, jira_search_by_jql, jira_xray_create_test
• ado_get_work_item, ado_move_to_state, ado_add_comment
• ado_list_prs, ado_get_pr, ado_add_pr_comment, ado_resolve_pr_thread, ado_merge_pr
• figma_get_layers, figma_get_icons, figma_download_node_image
• teams_send_message, teams_messages_since, teams_download_file
• gemini_ai_chat, openai_ai_chat, openai_ai_chat_with_files, bedrock_ai_chat

JavaScript Agent Pattern

All MCP tools are directly accessible as JavaScript functions in agents:

function action(params) {
    try {
        // Direct MCP tool access
        const ticket = jira_get_ticket(params.ticketKey);
        const analysis = gemini_ai_chat(`Analyze: ${ticket.fields.description}`);
        // Process and return
        return { success: true, result: analysis };
    } catch (error) {
        return { success: false, error: error.toString() };
    }
}

📚 Detailed Documentation