What is Chiron?

Chiron is a shared development environment where engineers and AI agents work from the same durable understanding of a codebase, its plan, and its history.

Most AI coding tools start every task from zero: they read a prompt, guess at context, and forget everything the moment the session ends. Chiron replaces that with a persistent, structured map of how your system actually works — code, documents, work items, decisions, and data — and hands both people and agents the same map. An agent that already knows your codebase does not need the codebase pasted into its prompt.

The three things Chiron gives an agent

• Knowledge — one graph across repositories, docs, tickets, commits, and connected databases, re-indexed as the branch moves. This is the Ontology Layer.
• Plan — intent expressed as Work Orders: the reasoning, the behavior, and the acceptance criteria, not a one-line ticket.
• Memory — what was learned on past work, surfaced at the right moment so the same context is never paid for twice. This is Memory.

An ontology, not an index

A search index tells an agent which files mention a thing. An ontology tells it what the thing *is* — what it is made of, what it depends on, and what breaks when it changes. Chiron types its entities and relations and shares meaning across sources, so "this code writes to that table" is a different, answerable fact from "this code reads it." Read Knowledge & the ontology for why that distinction is the whole product.

Deterministic first, models second

Structure — modules, symbols, imports, call paths — is derived by parsers, never by a model, and it is reproducible: the same commit produces the same graph every time. Models run *on top of* that structure and their output is labeled as interpretation. You always know whether a fact was computed or inferred.

Next: Getting started.