cheney369/Devora

What it solves

Devora provides a project context and engineering control plane for AI coding agents. It prevents the common issues where agents lack understanding of business boundaries, engineering rules, and historical decisions, which typically leads to inconsistent implementations and unsafe changes.

How it works

Devora acts as a layer on top of existing coding agents (such as Claude Code, Cursor, or Codex). It creates a .devora/ workspace within a project to maintain durable, agent-readable knowledge. It uses three core concepts:

• Governance: Defines project boundaries, architecture, and business constraints to make unknowns explicit.
• Skills: Encodes repeatable project-specific workflows into reusable capabilities for the agent.
• Change: Manages a specific piece of work from requirement to review, tracking scope, risks, and evidence.

Who it’s for

Developers using AI coding agents who need a way to ensure their agents maintain consistent project-specific knowledge and follow strict engineering and business rules across the development lifecycle.

Highlights

• Agent-Agnostic Integration: Works with multiple agents like Cursor, Claude Code, and OpenCode via specific adapters.
• Risk-Driven Quality Gates: Selects validation steps based on technical and business risks rather than relying solely on unit tests.
• Durable Context: Converts chat history and decisions into permanent project documentation in Markdown files.
• Controlled Scope: Tracks authorized file paths and prevents implementations from escaping their approved scope.