How versuscontrol/ai-infrastructure-agent works
Project type and purpose
• From the README and file tree, this is a full-stack Proof-of-Concept “AI Infrastructure Agent” that exposes a Web UI (dashboard) and a backend MCP Server which orchestrates an AI-driven AWS infrastructure manager.
• Key purpose: translate natural-language requests into safe AWS operations via AI planning, conflict detection, state management, and a visual web dashboard.
File structure
Top-level directories:
• cmd/web – Go entrypoint for the Web UI server (serves static assets + templates)
• cmd/server – Go entrypoint for the MCP Server (agent core + protocol server)
• web/ – React-style static assets (JS/CSS) and HTML templates under web/static & web/templates
• pkg/ – Core libraries, broken out by domain:
– pkg/agent: AI planning, request processing, execution, dependency resolution
– pkg/mcp: MCP protocol implementation (server, resource registry)
– pkg/aws & pkg/adapters: AWS SDK clients + adapter layer for each AWS resource
– pkg/state: Terraform-like state management & conflict detection
– pkg/discovery, pkg/graph, pkg/conflict: resource scanning, graph analysis, conflict resolution
– pkg/tools: helper functions for AWS resources
– pkg/web: optional shared HTTP handlers (if used)
• internal/: config loading and logging primitives
• scripts/: install and launcher helpers
Key insights from the README
• Components: Dashboard (forms + monitoring), Agent Core, AI Provider, AWS Client, State Management
• Safety features: Dry-run, conflict detection, drift management