TaskMaster AI Demo — Project Overview

Level: Intermediate · Advanced Pre-reading: 02 · Agentic AI · 04 · LangGraph · 05 · MCP Servers

This section documents a fully working demo of the AI-Powered Development Automation platform. A conversational chat agent accepts a JIRA ticket number, clones a GitHub repository, identifies the correct module to modify, generates and commits code changes, and opens a Pull Request — with human approval gates at every critical step.

What This Demo Builds

User types: "Work on TASK-42"
    ↓
Agent fetches JIRA ticket, classifies it (bug / story)
    ↓
Agent RAG-searches the codebase → identifies affected module
    ↓
Agent generates code diff + tests
    ↓
⏸ HUMAN GATE: "Here's the proposed change. Approve to proceed?"
    ↓
Agent commits to branch, pushes, opens GitHub PR
    ↓
Agent posts summary back to JIRA ticket
    ↓
User sees: "PR #17 created → https://github.com/..."

Demo Project: "TaskMaster"

A minimal but realistic multi-module Java project hosted on GitHub with three modules:

Two pre-configured JIRA tickets drive the demo:

End-to-End Architecture

graph TD
    User["👤 User·Chat UI"] -->|ticket number| ChatAPI["FastAPI<br/>Chat Engine<br/>/chat WebSocket"]
    ChatAPI --> LG["LangGraph<br/>Agent Orchestrator"]

    LG --> JIRA_MCP["JIRA MCP Server<br/>get_ticket · post_comment"]
    LG --> GH_MCP["GitHub MCP Server<br/>clone · push · create_pr"]
    LG --> RAG["RAG Service<br/>pgvector · Bedrock Embeddings"]
    LG --> LLM["AWS Bedrock<br/>Claude 3.5 Sonnet"]

    JIRA_MCP --> JIRA_API["JIRA Cloud<br/>REST API v3"]
    GH_MCP --> GH_API["GitHub REST API<br/>+ git operations"]
    RAG --> RDS["RDS PostgreSQL<br/>+ pgvector"]

    LG -->|interrupt| HumanGate["⏸ Human Approval Gate"]
    HumanGate --> User

    subgraph AWS["AWS Cloud"]
        ChatAPI
        LG
        JIRA_MCP
        GH_MCP
        RAG
        RDS
        LLM
        Secrets["Secrets Manager<br/>JIRA token · GitHub PAT"]
        S3["S3<br/>Code artefacts · Logs"]
    end

Cloud Services: AWS, GCP, Azure

AWS (Primary Example)

GCP Equivalent Stack

Azure Equivalent Stack

Cost Comparison (Estimated monthly for typical workload)

Recommendation: GCP is typically 15–20% cheaper for this workload. All three are suitable for production.

Documentation Index

Decision Guides (Start Here)

Multi-Cloud Implementation (No Vendor Lock-In)

Technical Deep-Dives (Cloud-Specific)

Prerequisites

Before starting:

• [ ] AWS account with Bedrock access enabled (Claude 3.5 Sonnet)
• [ ] Personal JIRA Cloud account (free tier is fine: https://yourname.atlassian.net)
• [ ] GitHub account with a new empty repository
• [ ] Docker Desktop installed locally
• [ ] Python 3.11+ and Node.js 20+ installed
• [ ] AWS CLI configured (aws configure)