What each tool actually does
Claude Code is a terminal-based AI agent that operates inside an existing codebase. You point it at a repo, describe a task, and it reads files, modifies code, runs tests, and iterates. The agent model is powerful for refactoring, debugging, adding features, and writing tests in a project that already has a structure. It does not generate fresh architectures, compliance documentation, or multi-tenant scaffolds from requirements.
Archiet starts earlier in the process. You describe what you're building (entities, user stories, compliance requirements, target stack). Archiet's Blueprint Wizard structures that into a formal architectural genome, and the generation pipeline emits a complete, production-grade codebase — backend, frontend, mobile, infrastructure, architecture documentation, and compliance docs — in one run. The output is a ZIP you download, boot with docker-compose up, and open in any editor or AI agent for customisation.
The comparison
| Factor | Claude Code | Archiet |
|---|---|---|
| Working inside an existing codebase | Excellent — reads the full repo, edits across files, runs tests | Not the use case |
| Generating a codebase from requirements | Can scaffold with prompts, but structure varies by session | Core use case: 12 stacks, deterministic pipeline, quality-scored output |
| Architecture documentation | None structured | ADRs (MADR v4) for every material technical decision under docs/decisions/ |
| Compliance documentation | None | GDPR DPIA, HIPAA risk assessment, PCI scope statement from your data model |
| Multi-tenant security defaults | Suggests what you ask | Built-in: workspace-scoped queries, httpOnly JWT, route guards on every protected endpoint |
| Supported output stacks | Whatever language is in the repo | Flask, FastAPI, Django, NestJS, Laravel, Go, Java, Rails, .NET, Salesforce Apex, SAP CAP, Tauri+Rust |
| Quality gate | No formal gate | ARCHVERIFY blocks output below 80/100 quality score |
| Database migrations | Can help write them if asked | Auto-generated Alembic migrations from entity model; never bare db.create_all() |
| Handoff documentation | None | Traceability matrix, security posture, cost TCO, handoff README in every ZIP |
| Pricing | Usage-based (Claude API tokens) | Free plan; Pro from $149/month |
The two tools operate at different phases. Archiet is day-zero: generate the complete project from your architecture. Claude Code is day-one-plus: work inside that project to add features, refactor, debug.
Why spec-driven generation matters
When Andrej Karpathy described the shift from "vibe coding" to spec-driven development in early 2026, the argument was practical: code generated from a formal specification is deterministic and auditable. When compliance requirements change, you change the spec and regenerate. When a new engineer joins, the spec documents why every decision was made.
Archiet's architectural genome is that spec. It captures entities, relationships, compliance overlays, technology choices, and acceptance criteria in a structured model. Generated code is derived from the genome. Claude Code then operates on the generated code — with the ADRs telling it why each architectural decision was made, and the OpenAPI spec defining the intended contract.
The combination is notably stronger than either tool alone. Claude Code working inside an Archiet-generated codebase has richer context: ADRs explain the decisions, the traceability matrix links requirements to code, and the OpenAPI spec defines the API contract. That context allows the agent to make changes that respect the architecture rather than drift away from it.
What Archiet generates
Every ZIP includes:
project/
├── app/ # Your chosen backend (Flask, FastAPI, Django, NestJS, etc.)
│ ├── blueprints/ # One Blueprint per entity
│ ├── models/ # SQLAlchemy ORM models + Alembic migrations
│ ├── services/ # Service layer; no business logic in routes
│ └── tasks/ # Celery tasks (if async jobs in your genome)
├── frontend/ # Next.js frontend with your screens
├── mobile/ # React Native / Expo mobile app
├── infrastructure/ # Docker, GitHub Actions, optional Terraform
├── docs/
│ ├── decisions/ # ADRs for every material technical decision
│ ├── traceability/ # Requirements → entities → routes → tests matrix
│ ├── compliance/ # DPIA, HIPAA, PCI docs (when flagged in genome)
│ └── security/posture.md # Attack surface, auth controls, pre-launch checklist
└── openapi.yaml # OpenAPI 3.1 spec (authoritative)
7 compliance frameworks automated: SOC 2, ISO 27001, GDPR, HIPAA, PCI-DSS, DORA, NIS2.
When to choose each
Choose Claude Code when: you already have a codebase and want an AI agent that can understand the full context of your project and execute complex multi-step tasks autonomously.
Choose Archiet when: you're starting a new project and want a complete, production-ready codebase generated from your requirements — with architecture documentation, compliance docs, and security defaults built in from day one.
Use both when: you want Archiet to generate the initial production scaffold on day zero, then use Claude Code (or Cursor, Windsurf, or any other AI coding agent) to add features and customise from day one.
FAQ
Does Archiet replace Claude Code?
No — they solve different problems. Archiet generates a complete project from an architecture spec. Claude Code works inside projects to implement tasks, refactor, and debug. The optimal workflow is Archiet on day zero for the initial generation, then Claude Code for ongoing development inside that codebase.
Can Claude Code generate a complete production app like Archiet?
Claude Code can create files and scaffold projects via prompts, but it works without formal architectural context. It doesn't generate database migrations, compliance controls, multi-tenant infrastructure, or architecture documentation in a structured, reproducible way. For a production application, you'd need to prompt Claude Code dozens of times and manually coordinate the results — Archiet does this in one generation from a formal spec.
Is the generated code compatible with Claude Code?
Yes. Archiet generates standard codebases in mainstream frameworks (Flask, FastAPI, NestJS, etc.). Claude Code reads any codebase. The ADRs and OpenAPI spec in every Archiet ZIP give Claude Code richer context than a typical scaffolded project.
What stacks does Archiet support?
Flask, FastAPI, Django, NestJS, Laravel, Go-Chi, Java Spring Boot, Rails, .NET, Salesforce Apex, SAP CAP, and Tauri+Rust — 12 backend stacks, all generating a Next.js frontend and Expo mobile app.
How long does Archiet generation take?
Typically 60-120 seconds for a complete blueprint.