Inspiration
Modern DevOps teams spend hours switching between logs, traces, dashboards, deployment records, source code repositories, and ticketing systems to diagnose production incidents. When a critical service slows down or starts failing, engineers must manually correlate information across multiple tools before identifying the root cause.
We built DevOpsGPT to automate this workflow using Agentic AI. Instead of manually investigating incidents, engineers can simply describe the problem in natural language, such as "Checkout API is slow", and DevOpsGPT performs the investigation autonomously. Our goal was to reduce incident response time, eliminate context switching, and demonstrate how AI agents can enhance observability and developer productivity.
What it does
DevOpsGPT is an autonomous engineering assistant that investigates production incidents using Splunk observability data.
A user submits a problem statement such as: Checkout API is slow The agent then:
- Searches Splunk logs for errors, warnings, and latency spikes.
- Analyzes traces and operational telemetry.
- Correlates anomalies with recent deployments.
- Identifies probable root causes.
- Generates remediation recommendations.
- Creates Jira tickets for incident tracking.
- Creates GitHub pull request drafts for proposed fixes.
The system provides a complete incident report containing findings, root cause analysis, severity assessment, and recommended actions.
How we built it
DevOpsGPT was built using a modular agent-based architecture.
Backend Python 3.11 FastAPI Uvicorn Pydantic AI Layer Anthropic Claude (primary reasoning engine) Provider-agnostic LLM abstraction Tool-calling agent architecture Observability Layer Splunk Enterprise Splunk REST APIs HTTP Event Collector (HEC) Real-time log ingestion Integrations Jira REST API GitHub REST API Architecture Flow User Query ↓ DevOpsGPT Agent ↓ Splunk Search & Analysis ↓ Root Cause Detection ↓ Recommended Fix ↓ Jira Ticket ↓ GitHub Pull Request
The application supports multiple operating modes, including live integrations and offline mock mode for demonstrations and testing.
Challenges we ran into
One of the biggest challenges was integrating multiple systems while maintaining reliability. Key challenges included:
Configuring Splunk REST authentication and management APIs. Connecting the agent to live Splunk data instead of mock datasets. Creating and validating HTTP Event Collector (HEC) tokens. Ingesting realistic operational telemetry into Splunk. Handling self-signed SSL certificates during local development. Designing a fault-tolerant architecture that gracefully falls back to mock services when external integrations are unavailable. Ensuring Claude could reason over operational data and generate meaningful remediation plans.
Accomplishments that we're proud of
We are proud that DevOpsGPT successfully demonstrates a complete agentic incident-management workflow. Highlights include: Successfully connected Claude to the investigation workflow. Integrated Splunk Enterprise using REST APIs. Configured HTTP Event Collector (HEC) for real-time log ingestion. Built a working web interface for natural language investigations. Created an autonomous root-cause-analysis pipeline. Implemented Jira and GitHub integration layers. Designed a resilient architecture with mock and live execution modes. Enabled end-to-end observability-driven investigations using real operational data. Most importantly, we transformed a traditionally manual troubleshooting process into a conversational AI experience.
What we learned
Throughout this project we gained hands-on experience with: Agentic AI architectures Tool-calling LLM workflows Splunk observability platforms Event ingestion through HEC Incident response automation API integrations and authentication FastAPI application development DevOps and Site Reliability Engineering practices We also learned that the true value of AI in DevOps is not simply generating responses, but orchestrating actions across multiple systems to accelerate decision-making and incident resolution.
What's next for DevOpsGPT:
Our roadmap includes:
Advanced Observability
- Native distributed tracing support
- Real-time anomaly detection
- Predictive incident forecasting ### Automated Remediation
- Automated rollback recommendations
- Infrastructure remediation workflows
- Runbook execution ### Enhanced Integrations
- Slack and Microsoft Teams notifications
- ServiceNow integration
- Kubernetes and cloud-native observability support ### AI Improvements
- Multi-agent collaboration
- Continuous learning from resolved incidents
- Context-aware remediation generation ### Enterprise Features
- Role-based access control
- Investigation history and audit trails
- Compliance and governance reporting Our long-term vision is to evolve DevOpsGPT into a fully autonomous AI-powered incident response platform capable of reducing mean time to detection (MTTD) and mean time to resolution (MTTR) across modern cloud environments. 🚀
Log in or sign up for Devpost to join the conversation.