Turn Claude Code into a structured development team

17 specialized agents. 6 phases from strategy to deployment. Handoffs through files, so context survives the session that created it. Start from an empty folder — or from the prototype you already shipped.

17
specialized agents
6
structured phases
MIT
licensed

Context gets lost. Quality varies. The same mistakes repeat.

A backend agent makes design decisions it shouldn't. Context disappears between conversations. There's no repeatable process.

You can build faster with AI. But only if the AI knows what it's working on, what it shouldn't touch, and where the last agent left off.

Built and refined across 10+ real projects. Every pain point became a fix.

Vibe coding gets you to a demo. It doesn't get you to a product.

That's not a knock — moving fast is the right call early on. But when an LLM hits a gap in context, it doesn't stop. It fills the gap with whatever looks right, and those guesses compound. Archflow closes the gaps before code is written, so what you build holds together as it grows.

Vibe Coding vs Archflow — comparison showing how vibe coding loops through gaps and hallucinations while Archflow uses structured phases with human approval gates

One agent per job. Files as handoffs. Phases as guardrails.

Specialized Agents

17 agents, each scoped to one domain. A UX designer doesn't write backend code. An API engineer doesn't make design decisions. Deep expertise, not broad awareness.

File-Based Handoffs

Agents communicate through artifacts, not chat. The API contract becomes the single source of truth. Frontend and backend build against the same spec. Nothing gets lost between conversations.

Phase Gates

6 phases from strategy to deployment. Nothing moves forward without your approval. No skipped steps. No autonomous decisions on what ships.

From idea to production in 6 phases

1

Strategy & Planning

product-strategist / feature-planner

2

Design

ux-designer / dsl-generator

2.5

API Architecture

api-contract-architect

3

Implementation (parallel)

ui-engineer + api-engineer / qa-engineer

4

Quality & Optimization

code-reviewer / performance-optimizer

5

Launch & Operations

devops-engineer / post-launch-analyst

Each phase loads only what the active agents need. Less noise. More focused results.

Already shipped something? Start from there.

Your prototype works. People are using it. What it doesn't have is a roadmap, an API contract, a test plan, or a defensible answer to what 1.0 means.

/archflow:onboard

Onboarding audits what you shipped and reconstructs the plan around it. Where it finds working code with no story behind it, it writes the story and marks it delivered — so your roadmap opens describing reality, not pretending you're on day one.

Then close the gaps that are costing you

Can't decide what to build next

Phase 1 — strategy, personas, KPIs

Every screen looks slightly different

Phase 2 — design system

Frontend and backend disagree about the API

Phase 2.5 — contract both sides build against

Scared to change anything

Phase 3 — tests and acceptance criteria

No idea if it's secure or fast enough

Phase 4 — review and optimization

No repeatable way to ship

Phase 5 — CI/CD, versioning, app store

What this does and doesn't do: you get the plan, contracts, and quality gates your prototype never had, and Phase 4 agents review code quality, security, and performance. Archflow will not silently rewrite your architecture. You decide what gets refactored — it makes sure you know what needs it.

Full guide: taking a prototype to production

Inside /archflow:onboard

Most AI workflows assume you're starting from scratch. This is the step that doesn't — here's exactly what it does to your repo.

Run /archflow:onboard and it dispatches up to 9 agents in parallel. They audit your code, import docs from Jira, Notion, Linear, GitHub, and Slack, reverse-engineer your design system and API contracts, and drop you into the right phase.

1

Answer 5 questions

2

Up to 9 agents analyze in parallel

3

Artifacts generated for your approval

New project? /archflow:init starts you at Phase 1.

Too much process for a weekend project?

Then don't run it. Archflow has two modes over the same schema, and new projects start in the light one.

quick

Solo, small, moving fast

One implicit release. Gates that record but never block. No ship ceremony until you ask for one. You still get agents, contracts, and phases — without the paperwork.

full

Team, releases, real stakes

Explicit release pipeline, enforced readiness gates, role lanes for PM, design, architecture, and QA. Everything a second contributor makes necessary.

Same schema either way, so switching costs nothing. When Archflow notices you've outgrown quick — a second release, a second contributor — it offers the upgrade. It never forces it.

Three commands, most days.

There are eight in total, but this is what the loop actually looks like.

/archflow Where the project stands, and what to run next
/archflow:feature Add a story — capture it, or branch and build it now
/archflow:groom Turn a backlog stub into a ready story with acceptance criteria

Plus, when you need them

/archflow:init Start a new project at Phase 1
/archflow:onboard Analyze an existing codebase
/archflow:release Create, start, and ship releases
/archflow:mode Switch between quick and full
/archflow:setup-mcp Connect Jira, Notion, Linear, GitHub
/archflow:migrate Upgrade a v1.0 project to v2.0

Adapts to what you're building

Fullstack

Frontend

Backend

Mobile

Phases, agents, and roadmap structure adjust based on project type. Set automatically during onboarding.

Two commands to install

$ claude plugin marketplace add archflow https://github.com/AZidan/archflow
$ claude plugin install archflow --scope project

Free and open source. No lock-in. Uninstall anytime.