echopath-labs/openDomain

What it solves

OpenDomain provides a way to maintain a "semantic layer" for business logic that is readable by both humans and AI agents. It prevents AI agents from hallucinating or arbitrarily changing business rules by ensuring that long-lived concepts, rules, and lifecycles are stored as evidence-backed Markdown files in a Git repository, requiring human approval before any AI-inferred knowledge becomes official.

How it works

The system uses Markdown files with YAML front matter as the source of truth. AI agents (specifically Codex) can propose new business knowledge as "Candidates." These candidates are kept separate from accepted knowledge and must be explicitly reviewed and approved by a human maintainer. The tool includes a CLI for initialization, validation of reference integrity, and exporting context for agents to use during implementation tasks, ensuring that agents are "grounded" in the accepted business domain before they modify code.

Who it’s for

It is designed for software maintainers and AI agents working together on complex business systems where maintaining a strict, human-verified record of business rules and domain semantics is critical.

Highlights

• Git-native storage: Business concepts and rules are stored in repository-readable Markdown.
• Human-in-the-loop: AI-inferred knowledge starts as a Candidate and requires explicit human review to be accepted.
• Reference Integrity: Includes schema validation and integrity checks to ensure the domain model is consistent.
• Agent Grounding: Provides a protocol to ensure AI agents are grounded in the correct business context before performing tasks.
• Package-neutral installation: Can be installed via npm or as a standalone binary without modifying the host project's metadata.