Aurora is an open-source AI-powered incident management platform designed for SRE teams, automating the complex process of incident investigation and root cause analysis (RCA). It empowers teams by autonomously querying across multi-cloud environments (AWS, Azure, GCP, etc.) and Kubernetes, correlating data from numerous tools, and generating actionable RCAs with remediation suggestions, thereby shifting focus from workflow automation to investigation automation.

How It Works

Aurora employs LangGraph-orchestrated AI agents that dynamically select and utilize over 30 integrated tools to investigate incidents. These agents execute commands (e.g., kubectl, aws, az, gcloud) within sandboxed Kubernetes pods, query logs from various sources, analyze recent deployments, and traverse an infrastructure dependency graph powered by Memgraph. This approach allows for deep, autonomous investigation across disparate systems, enabling faster and more accurate root cause identification compared to traditional, workflow-centric automation tools.

Quick Start & Requirements

Local setup involves cloning the repository, initializing configuration (make init), and adding an LLM API key (e.g., from OpenRouter or OpenAI) to the .env file. Starting Aurora is achieved via make prod-prebuilt or make prod-local. A Vault root token is also required and obtained from container logs, then added to .env. The only external requirement for basic functionality is an LLM API key; cloud provider accounts are optional. Aurora is accessible via http://localhost:3000.

• Docs: https://aurora.arvo.ai/docs
• Website: https://aurora.arvo.ai/

Highlighted Details

• Agentic Investigation: AI agents autonomously select and use 30+ tools for deep investigation across cloud and Kubernetes environments.
• Multi-Cloud & Hybrid Support: Integrates with AWS, Azure, GCP, OVH, Scaleway, Cloudflare, and Kubernetes.
• Flexible LLM Integration: Supports OpenAI, Anthropic, Google Gemini, Vertex AI, OpenRouter, and local models via Ollama for air-gapped deployments.
• Automated RCA & Postmortems: Generates structured RCAs, impact assessments, remediation steps, and suggests code fixes, with export options to Confluence and Notion.
• Self-Hosted & Free: Fully self-hostable via Docker Compose or Helm, with no per-seat or per-incident pricing.

Maintenance & Community

Aurora is an active open-source project with community support channels via Discord, GitHub Issues for bug reports/feature requests, and GitHub Discussions for general conversation. Contributions are welcomed following established guidelines.

• Discord: https://discord.gg/arvo-ai
• GitHub: https://github.com/arvo-ai/aurora

Licensing & Compatibility

Aurora is licensed under the Apache License 2.0, permitting commercial use and modification, with standard attribution requirements.

Limitations & Caveats

While core functionality requires only an LLM API key, enabling cloud provider integrations necessitates configuring specific cloud credentials. The self-hosted nature implies the user is responsible for managing the underlying infrastructure (Docker or Kubernetes). Network access is required for LLM API calls and Terraform registry lookups unless using local Ollama models in an air-gapped setup.