NexTOps Agent

NexTOps is an AI-driven DevOps solution that works on Jira stories with Atlassian, Terraform, GitHub, K8S agents. The React UI is created to see the response streamed by the Bedrock Agentcore Runtime.

Comment

NexTOps — AI-Powered DevOps Agentic Solution

Inspiration

As a DevOps architect, We have noticed that much of daily work involves repetitive tasks: creating infrastructure, writing Terraform scripts, managing Jira stories, and setting up CI/CD pipelines. We wanted to build an AI-driven solution that could handle these end-to-end tasks autonomously, saving time and reducing human error.

The AWS Global Hackathon provided the perfect opportunity to demonstrate intelligent agents that collaborate across tools like Jira, GitHub, Terraform, EKS, and MCP-managed servers.

What We Learned

  • How to integrate multiple cloud services and DevOps tools with AI agents.
  • Techniques for building agentic systems using AWS Bedrock AgentCore, including short-term and long-term memory for workflow management.
  • Challenges of orchestrating multi-agent workflows, handling dependencies, and ensuring reliable task completion.
  • Practical lessons in automating GitHub workflows, Terraform script generation, and CodePipeline setup.
  • The role of MCP servers in providing secure, standardized connectors for Jira, GitHub, Terraform, and EKS.

How We Built It

  1. Architecture:

  * Frontend: React UI integrated with Cognito and IAM for secure login.   * Backend: API Gateway & Lambda for invoking agents and showing real-time workflow progress.   * Agents:

    * Jira Agent: Pulls task details from Jira and updates tickets via MCP connectors.     * Terraform Agent: Reads docs, generates reusable modules, and applies infrastructure changes through MCP-managed Terraform servers.     * GitHub Agent: Creates feature branches, pushes scripts, and sets up CI/CD workflows using MCP connectors for GitHub and AWS services.   * Bedrock AgentCore powers the agents with STM & LTM memory and a knowledge base.

  1. Workflow:

  * Assign a Jira ID → Jira Agent fetches details via MCP → Terraform Agent generates scripts → GitHub Agent pushes code → Jira Agent updates ticket.   * Reusable modules from one task are leveraged for future tasks to provision ECR, EKS, and CI/CD pipelines.

  1. Automation Highlights:

  * End-to-end DevOps automation for dev, test, and prod environments.   * CI/CD pipeline creation for application onboarding.   * Autonomous error handling: agents can fix issues on request.   * MCP connectors provide seamless integration between Jira, GitHub, Terraform, and EKS, ensuring secure and reliable automation.

Challenges Faced

  • Multi-agent coordination: Ensuring Terraform, GitHub, and Jira agents work in sync without conflicts.
  • Error propagation: Handling failures in one agent without breaking the entire workflow.
  • Knowledge extraction: Parsing documentation to generate reusable Terraform modules reliably.
  • Real-time UI updates: Displaying agent progress and conversations smoothly in React.
  • MCP integration: Ensuring each connector communicates securely and consistently with the corresponding service. Especially the Jira MCP was very hard to integrate.

Reflection

Building NexTOps taught me how AI agents can transform DevOps operations. By combining automation, cloud infrastructure, MCP connectors, and agent collaboration, We were able to create a system that handles end-to-end workflows with minimal human intervention.

This project reinforced the power of agentic AI in cloud-native environments, the importance of modular design, robust orchestration, and secure integration using MCP servers.

Built With

Share this project:

Updates