What is Knowledge Tree

The Problem

Modern infrastructure spans multiple cloud providers, Kubernetes clusters, DNS zones, and network segments. Teams lose visibility into what exists, how components depend on each other, and what documentation is outdated. Onboarding new engineers takes weeks. Incident response requires tribal knowledge. Compliance audits are manual and error-prone.

The Solution

Knowledge Tree solves this by continuously discovering infrastructure resources across AWS, Azure, GCP, and Kubernetes, storing them in a knowledge graph with their relationships, enriching them with AI-generated summaries, and producing always-current documentation automatically.

Core Capabilities

• Multi-cloud discovery — Plugins for AWS, Azure, GCP, Kubernetes, DNS, and network scanning. Extensible via a Go plugin SDK.
• Knowledge graph — PostgreSQL with Apache AGE for Cypher queries and pgvector for semantic search. Every resource is a node, every dependency is an edge.
• AI enrichment — Python-based enricher generates service descriptions, runbooks, security analyses, and executive summaries using local or cloud LLMs.
• Auto-documentation — Generates Markdown, HTML, and Mermaid diagrams. Publishes to Confluence and keeps them in sync.
• Change detection — Snapshots infrastructure state and diffs against previous versions to track drift.
• FinOps — Cost intelligence API estimates monthly and annual spend by provider, category, and resource type.
• Compliance — Built-in rules evaluate resources for security, cost, reliability, and governance issues with an overall health score.
• AI integration — Model Context Protocol (MCP) server lets AI agents query the graph directly. Chat API for natural-language exploration.

How It Works

1. Discover — Plugins scan your infrastructure on a configurable schedule (default: every 15 minutes). Each plugin emits resources and relationships.
2. Store — Resources are upserted into the knowledge graph. Relationships are created as directed edges. Vector embeddings are generated for semantic search.
3. Enrich — The Python enricher service calls configured LLMs to generate descriptions, runbooks, and security analyses for each resource.
4. Document — Auto-generated documentation is written to disk or published to Confluence. Documentation refreshes when the graph changes.
5. Query — Use the REST API, Cypher queries, or the chat interface to explore your infrastructure. AI agents connect via MCP.

Technology Stack