griddynamics/rosetta

What it solves

Rosetta addresses the problem of fragmented AI coding agent usage across teams. Instead of individual developers writing their own prompts or agents guessing project architecture from a few open files, Rosetta provides a shared, versioned system for engineering governance and context. It prevents agents from confidently making wrong decisions by encoding team-specific architecture, standards, and workflows into every session.

How it works

Rosetta acts as a runtime layer that orchestrates instructions across three levels: Core (best practices), Organization (standards and policies), and Project (architecture and constraints). These layered instructions are then passed through agent adapters to tools like Claude Code, Cursor, Copilot, and Codex. Everything is authored in Markdown and versioned in Git, allowing instructions to be portable across different IDEs and agents.

Who it’s for

It is designed for engineering teams and organizations that use AI coding agents and want to maintain consistent engineering discipline, shared standards, and governed workflows across multiple repositories and developers.

Highlights

• Reusable Skills: A library of on-demand units of work covering discovery, requirements authoring, design, coding, testing, security, and orchestration.
• Structured Workflows: Pre-defined end-to-end flows for tasks like coding-flow, security-flow, and testgen-flow that include approval gates.
• Multi-Agent Orchestration: Ability to brief subagent teams and use fresh-context reviewers to avoid self-review bias.
• Tool Agnostic: Works with existing MCP-compatible agents and IDEs, preventing vendor lock-in.
• Governance: Supports versioned control via Git and can run within a secure perimeter.