Introduction

Ontoly turns a TypeScript repository into a deterministic Software Graph.

Most developer tools keep rebuilding the same partial understanding of a codebase. Agents search files. Documentation tools parse declarations. SDK generators reconstruct schemas. Architecture tools rebuild dependency graphs. Ontoly gives those tools one shared semantic layer.

The product is the graph.

Ontoly does not answer questions, generate code, or reason with AI. It discovers files, parses source, resolves symbols, extracts relationships, builds indexes, and persists a graph that other tools can consume.

Where to start

• Installation - add the package and initialize a repository.
• Build your first graph - run the compiler and inspect artifacts.
• Query the graph - use the deterministic query engine.
• Semantic Index - resolve natural software concepts to graph entities without AI.
• Semantic Intelligence - derive feature ownership, intent vocabulary, neighborhoods, and concept graphs.
• Repository Intelligence - derive Git-backed ownership, hotspots, co-changes, churn, and architectural drift.
• Software Graph - understand the node and relationship model.
• Capabilities - deterministic software-engineering answers over the graph.
• Capability API - call the high-level capability engine from code, CLI, or MCP.
• Enhancers - build deterministic artifacts above the immutable Software Graph.
• Evidence Model - understand how answers cite graph facts.
• Confidence Model - see how confidence is derived from graph evidence.
• LLM Enhancement - mandatory workflow rules when an LLM uses Ontoly.
• TypeScript Semantic Model - inspect the pure language model.
• Framework Analyzer API - add deterministic framework analyzers.
• Semantic Generator - understand how semantic facts become graph artifacts.
• Plugin system - extend Ontoly without changing the compiler.
• Semantic Coverage - measure graph completeness, confidence, and trustworthiness.
• Semantic Evaluation Harness - gate releases on deterministic software-understanding questions.
• Validation Lab - continuously measure correctness, determinism, performance, and regression resistance.
• Agent Skills - install workflow-only skills that teach agents to use Ontoly first.
• Skills Overview - browse the official skill and capability matrix.
• Agent Skills Catalog - browse public docs pages for every installable Ontoly Skill.
• Skills Validation - validate skill structure, references, examples, and agent behavior.
• FAQ - common first-time user questions.
• Troubleshooting - actionable fixes for common failures.
• Contributing - project workflow and release-gate expectations.
• NestJS Support - inspect framework-specific graph semantics.

Release references

• Historical Alpha Release Notes - what shipped before the Release Candidate freeze.
• Known Limitations - what the Release Candidate intentionally does not cover.
• Compatibility Matrix - runtime, graph, skill, and MCP compatibility.
• Feature Matrix - supported, Release Candidate, and non-goal surfaces.
• Framework Matrix - current framework validation status.
• Version Matrix - package and spec versions.

Boundaries

• Not a chat interface.
• Not an AI agent.
• Not vector search.
• Not hosted SaaS.
• Not code generation.

Plugins may build AI-facing tools, SDK generators, docs, diagrams, or reports. The compiler remains deterministic and AI-free.

Status

Release Candidate. The TypeScript frontend, Software Graph, Query Engine, MCP, Skills, validation lab, and benchmarks are in v1 freeze. Changes that affect the graph contract, public APIs, packages, or skill compatibility still require the RFC and release-gate process.