Gator is a Git-native semantic layer for AI-assisted engineering.
It gives coding agents a durable, reviewable map of the repository: its modules, responsibilities, relationships, invariants, and non-obvious constraints. Gator calls these maps charters.
Before changing code, an agent reads the relevant charters to understand the part of the system it is about to touch. As it works, Gator requires the agent to update those charters alongside the code. Pre-commit Git hooks check that the knowledge layer and the implementation stay in step.
This means an agent does not need to rediscover the entire repository from raw files every session. It can begin with a structured understanding of the system and progressively inspect the details that matter for the task.
The experience should feel less like prompting a stranger with a large folder of code, and more like working with a senior engineer who already knows the system: what each part owns, what depends on it, and what must not break.
Gator also provides a repository constitution: shared instructions for how agents work in this codebase. The constitution defines the change process—read the relevant charters, propose a plan, make the change, update the map, record the decision, and commit through Git’s normal controls.
The result is a living knowledge layer that travels with the repository. Git records not only what changed, but the maintained understanding that guided the change.