Context

Who is this for? Contributors, architects, and stakeholders who need to understand the business purpose, domain, and requirements behind Rosetta.

When should I read this? When you need to understand why Rosetta exists, who it serves, and what success looks like. Read the Introduction first. For technical details, see Architecture.

What this document is. Business context, stakeholder perspective, and target state. Bulleted, concise, no technical implementation details. Source of truth for project purpose and business requirements.

What this document is not. Not architecture, not a user guide, not a changelog.

Why Rosetta Exists

• Every engineer writes their own prompts, rules, and guardrails (or none at all)
• AI agents lack business context, conventions, architecture knowledge, and compliance requirements
• Knowledge stays trapped in silos and senior engineers’ heads
• No visibility, no consistency, no enforcement of AI behavior at scale
• Low adoption rates due to compliance exposure and reinvented wheels
• Expertise walks out the door with no way to capture or transfer it

Result: Slow AI adoption, compliance risk, duplicated effort, lost institutional knowledge.

What Rosetta Is

Rosetta is a meta-prompting, context engineering, and centralized instructions management for AI coding agents — an open-source governance and context layer. It provides structured context - rules, skills, workflows, and sub-agents - guiding AI systems to operate with a deep understanding of system architecture, domain constraints, and engineering standards. Rosetta also accelerates project onboarding by reverse-engineering architecture and domain context, improving the reliability and consistency of AI-generated code.

It comes preloaded with battle-tested best practices from real-world projects.

Not reactive like gateways. Not static like prompt libraries. Verified, project-specific, and tool-agnostic.

Value Delivery

For Engineers

• Standardized, expert-prepared workflows with batteries included
• Quick onboarding with automatic synchronization of best practices
• Reduced need for prompt engineering expertise
• Universal compatibility across any coding agent and tech stack
• Human-in-the-loop support built into workflows

For Managers

• Consistent instruction setup across the organization
• Built-in guardrails reduce AI risks and rework
• Complete workflows cover steps that junior engineers forget
• Optimized token consumption across workflows
• Team mobility across projects without performance loss

For Directors

• Adoption and usage tracking with per-feature visibility (opt-in, self-hosted MCP — see MCP Context)
• Department-wide governance with centralized management
• Always-current technology information and curated business knowledge
• A/B testing capability for SDLC experimentation
• Risk mitigation through secure internal deployment

For VPs and Executive Leadership

• Measurable outcomes with reduced implementation risk
• Governance and compliance coverage
• Transparent adoption and usage metrics (opt-in, self-hosted MCP — see MCP Context)
• Cost efficiency when introducing AI to the organization
• IP protection preventing intellectual property leakage
• Future-proofing: adapts to new models and tools with minimal effort

Business Requirements

Speed

• Onboarding a new repository: minutes, not weeks (target: 15 minutes vs. 2 weeks manual setup)
• Coding task preparation: copy a story and review a spec, not 60 minutes of manual prompt crafting
• Instruction updates: author, review, publish, and roll out without downtime
• Publishing: change detection ensures only modified instructions are pushed (seconds, not full republish)

Simplicity

• One-command installation for any supported IDE
• No local dependencies beyond a standard development environment
• Progressive disclosure: agents load only what they need, when they need it
• Works with existing tools and workflows, no process overhaul required

Scale

• Same instructions, same behavior across all teams and projects
• Organization-wide rollout from a single instructions repository
• Release-based versioning (r1, r2, r3) for safe evolution and rollback
• Layered customization: core baseline + organization overrides + project-specific tweaks
• Plugin distribution via IDE marketplaces (Claude Code, Cursor) for instant installation across projects. See Installation — Plugin-Based Installation

Governance

• Instructions are versioned, reviewed, and approved like code (rules-as-code)
• Built-in guardrails, risk assessment, and human-in-the-loop checkpoints
• Adoption tracking and usage analytics per feature, team, and project
• Runs inside the organization’s perimeter (works with limited internet access)

Quality

• Preloaded with proven patterns from real-world engagements
• Evaluation and judge pipelines for instruction quality
• Consistent agent behavior reduces hallucinations and rework
• Observed time savings per coding task: with Rosetta, ~5 min to type, ~5 min to review, ~15 min for AI to execute (25 min total). Without Rosetta, ~30 min to type, ~15 min back-and-forth in planning mode, ~15 min to execute, ~15 min to catch up (75 min total). In practice, we see productivity scale to 3x-5x, though individual results vary by task complexity

Domain and Operating Context

• Rosetta operates across the full Software Development Lifecycle: planning, requirements, implementation, QA, release, and operations
• Agent-agnostic: works with Cursor, Claude Code, VS Code, Windsurf, JetBrains (Copilot, Junie), GitHub Copilot, Codex, Antigravity, OpenCode, and any MCP-compatible IDE — via a plugin where supported, via MCP otherwise. Native hook support is IDE-specific; OpenCode hooks are not currently supported or tested.
• Security model: Rosetta is designed to only serve knowledge and instructions to agents — it does not see or process source code. No code leaves the organization’s perimeter by design. Rosetta itself does not store any project data. Self-hosted MCP deployments can opt into additional features (cross-project intelligence, execution tracking, adoption analytics) — see MCP Context for what those involve and who’s responsible for the data. None of them exist in plugin mode.
• Grounded in production experience, not theory. Active feedback loop from real projects.

Target State

• Every AI coding agent in the organization receives consistent, expert-prepared instructions from day one
• Knowledge is versioned, searchable, and automatically synchronized
• Best practices evolve safely with rollback capability
• Business context flows to agents on demand through Rosetta’s structured instruction delivery, not manual copy-paste
• Organizations govern AI behavior with the same rigor they govern code
• Instruction authoring, review, and publishing is a standard engineering workflow

Edition

• Rosetta: Fully open-source, complete source code, self-hosted, community-driven. Includes all public agents, workflows, skills, rules, and templates.

For a comparison of installation modes (Plugin, Offline, and the optional MCP transports HTTP and STDIO), see the Installation — Choose Your Mode table.

Design Philosophy

These principles shape every product and architectural decision:

• Agent-agnostic. Never lock users into one IDE or AI model.
• Progressive disclosure. Load instructions in stages. Prevent context overflow. Give agents only what the current task requires.
• Classification-first. Every request is auto-classified before work begins. Classification drives what loads. See Usage Guide — Workflows for available request types.
• Release-based versioning. Develop new instructions without breaking agents on stable versions.
• Rules-as-code. AI behavior is authored, versioned, reviewed, and approved through standard engineering workflows.
• Security by design. No source code transfer. Works with limited internet access. Runs inside the organization’s perimeter.
• Batteries included. Ship proven defaults. Make the right thing the easy thing.

Related Docs

• Introduction: What Rosetta is and where to start
• Overview: Mental model, terminology, design principles
• Architecture: System structure, components, data flow
• MCP Context: Business case, data responsibility, and opt-in features for self-hosted MCP (optional, secondary)
• Developer Guide: Repo navigation, where to change what
• Contributing: Fastest path to merged PR
• Review Standards: What reviewers verify, what authors provide
• Quick Start: Zero to working setup
• Usage Guide: How to use Rosetta day-to-day
• Deployment: Self-hosted MCP server and infrastructure setup (optional, rarely needed)
• Troubleshooting: Symptom-first diagnosis
• Installation: Full setup reference for complex environments