OpenClaude Workflow
vinileonardo/openclaude-workflow
Orquestração de 10 agentes IA especializados para dev completo
Ver no GitHubSistema profissional de múltiplos agentes IA para desenvolvimento com Claude Code. Orquestra 10 agentes especializados com sistema adaptativo automático, pilha de modelos por complexidade e memória estruturada entre sessões. Suporta projetos verdes (greenfield) e integração em projetos existentes.
Instalação
npx convez-ai vinileonardo/openclaude-workflow1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Automatizar pipeline de spec até deploy com agentes especializados
- Integrar workflow IA em projetos existentes com memory bootstrapping
- Aplicar quality gates (revisão de código, segurança, testes) de forma estruturada
- Adaptar setup de desenvolvimento automaticamente à stack do projeto
Veredito

Ferramenta robusta para times que já trabalham com Claude Code e buscam orquestração industrial de agentes. Curva de aprendizado alta, mas oferece estrutura completa de CI/CD com IA. Recomendado para projetos com equipe técnica experiente.
README
A tradução para o português deste README ainda está na fila.
OpenClaude Workflow
A professional multi-agent AI development workflow for OpenClaude (the open-source CLI for AI-assisted coding). Drop-in setup for any project — greenfield or existing.
Quick Start
# 1. Install OpenClaude
npm install -g @anthropic-ai/claude-code
# Or download from: https://github.com/Gitlawb/openclaude
# 2. Install OpenCode Go (API provider)
# See: https://github.com/opencode-ai/opencode-go
# 3. Add the alias to ~/.bashrc
alias openclaude='OPENAI_BASE_URL=https://opencode.ai/zen/go/v1 \
CLAUDE_CODE_USE_OPENAI=1 \
OPENCODE_API_KEY=sk-your-key-here \
OPENAI_MODEL=deepseek-v4-flash \
openclaude --dangerously-skip-permissions'
source ~/.bashrc
# 4. Start OpenClaude in your project
cd your-project
opencode
# 5a. New project? Run the greenfield setup
/init-workflow
# 5b. Existing project with code already written? Run adoption
/init-workflow --adopt
The skill analyzes your stack, commands, and architecture, then generates a complete agent ecosystem adapted to your project.
Two Setup Modes
| Mode | Command | Best For |
|---|---|---|
| Greenfield | /init-workflow | New projects or no existing workflow setup |
| Adoption | /init-workflow --adopt | Existing projects with code, conventions, and possible existing workflow |
Greenfield (/init-workflow)
- Discovery — reads your project (stack, commands, structure)
- Generation — creates adapted agent files, CLAUDE.md, AGENTS.md, memory
- Confirmation — shows what was detected, asks for approval
- Guidance — explains how to use the workflow
Adoption (/init-workflow --adopt)
For projects that already have code and possibly a partial setup:
- Inventory — scans existing workflow files, configs, and git history
- Memory Bootstrapping — extracts coding conventions and patterns from existing code
- Diff Plan — previews what will be created, overwritten, or preserved (use
--dry-runto preview only) - Merge & Write — merges templates with your existing content (preserves custom agents, rules, memories)
- Adoption Report — summary of what changed and next steps
# Preview adoption changes without writing
/init-workflow --adopt --dry-run
# Quick inventory only
/init-workflow --check
Features
- 10 Specialized Agents — From spec to deploy, each with a clear role and model
- Native OpenCode Go Models — deepseek-v4-pro, qwen3.7-plus, deepseek-v4-flash — mapped to agents by task complexity
- Full Orchestration Pipeline — spec-driver → product-owner → explorer → devs → test-writer → code-reviewer + doc-writer → security-reviewer → devops
- Auto-Adaptation —
/init-workflowreads your project and generates files with correct paths and commands - Adoption Mode —
/init-workflow --adoptbootstraps memory from existing code and merges with your current setup - Structured Memory System — Persistent cross-session context with typed memories (user, feedback, project, reference)
- Reviewer Decision Trees — Clear rules for blocking vs. non-blocking feedback
- Quality Gates — Code review before security review before merge
Agent Workflow
spec-driver → product-owner → explorer → devs → test-writer → code-reviewer
+ doc-writer (parallel)
↓
security-reviewer
↓
devops
Agent Roles
| Agent | Model | Purpose |
|---|---|---|
| spec-driver | deepseek-v4-pro | SDD specs from vague requests |
| product-owner | deepseek-v4-pro | Issues, backlog, priorities |
| explorer | deepseek-v4-flash | Codebase mapping, research |
| backend-dev | qwen3.7-plus | API, logic, database |
| frontend-dev | qwen3.7-plus | UI, components, state |
| test-writer | qwen3.7-plus | Unit/integration tests |
| code-reviewer | deepseek-v4-pro | Quality, bugs, standards |
| security-reviewer | deepseek-v4-pro | OWASP, audit, compliance |
| documentation-writer | deepseek-v4-flash | Docs, changelog, ADRs |
| devops | qwen3.7-plus | Docker, CI/CD, deploy |
Model Guide
| Model | Provider | Speed | Best For | Agents |
|---|---|---|---|---|
| deepseek-v4-flash | DeepSeek | Fastest | Read-only, scanning, simple output | Explorer, Doc Writer |
| qwen3.7-plus | Qwen | Fast | Implementation, commands, testing | Backend/Frontend/Test/DevOps |
| deepseek-v4-pro | DeepSeek | Slower | Deep analysis, reviews, decisions | Spec/PO/Code/Security Review |
Project Structure
openclaude-workflow/
├── template/ # → Copy into target project
│ ├── CLAUDE.md # Pointer with placeholders
│ ├── AGENTS.md # Orchestration workflow
│ ├── .claude/
│ │ ├── settings.json # OpenClaude config
│ │ └── agents/ # 10 agent definitions
│ └── memory/
│ ├── MEMORY.md # Memory index template
│ └── _examples/ # Example memories
├── skills/
│ └── init-workflow/ # /init-workflow skill
│ ├── SKILL.md # Skill instructions
│ └── references/ # Detection patterns
├── docs/
│ ├── SETUP.md # Setup guide
│ ├── SKILL_GUIDE.md # Full skill reference
│ ├── CUSTOMIZATION.md # Adaptation guide
│ └── AGENT_TIERS.md # Model philosophy
└── archive/
└── oc-go-cc-config.example.json # Archived proxy reference
Documentation
| File | Description |
|---|---|
AI_ENTRYPOINT.md | Leitura obrigatória para IAs — fluxo de decisão ao ser direcionada ao repo |
docs/SETUP.md | Setup guide with prerequisites and troubleshooting |
docs/SKILL_GUIDE.md | Complete reference for the /init-workflow skill |
docs/CUSTOMIZATION.md | How to add/modify agents, models, and project rules |
docs/AGENT_TIERS.md | Model philosophy and tier strategy |
License
MIT