The product direction
Arth.sh is being built as an agnostic repository-to-production hub: an AI-native control plane for understanding, changing, verifying, deploying, and operating software without forcing teams into one framework, cloud, runtime, or hosting platform.
The product begins with the repository rather than a generated sandbox. It is designed to import an existing codebase, understand how the system is structured, and retain project memory across planning, implementation, review, deployment, and operation.
The problem
Current AI development tools are often excellent at producing a screen or a small application, but become less dependable when software already has history, multiple services, deployment constraints, tests, data contracts, infrastructure, and human review requirements.
Arth addresses that gap by treating software delivery as a controlled engineering loop rather than a one-shot generation task.
What Rokad is building
- Repository and architecture understanding across supported stacks.
- Change planning with dependency and impact analysis.
- AI-assisted implementation that remains visible and reviewable.
- Verification through tests, checks, previews, and deployment evidence.
- Git-native branches, commits, pull requests, and reversible changes.
- Deployment adapters for multiple platforms and container targets.
- Production feedback through logs, traces, incidents, and project memory.
- Human approval at the points where risk or ownership requires it.
Product principles
Agnostic by architecture
Frameworks, runtimes, clouds, and deployment providers are adapters around the control plane—not assumptions embedded into the product.
Repository-native
The repository remains the source of truth. Arth works with existing engineering practices instead of hiding changes inside an isolated proprietary workspace.
Production-oriented
Planning, testing, deployment, rollback, observability, and operational context are treated as part of software creation from the beginning.
Human-controlled
AI increases engineering throughput, but the user retains control over decisions, changes, approvals, and releases.
Current stage
Arth is in active development. The first production vertical slice covers repository import, system understanding, an AI-assisted engineering loop, verification, Git synchronization, preview deployment, and controlled release foundations.