create-opencode-workflow
v4.36.27
Published
Initialize OpenCode Workflow - AI-assisted development system. Alias for @comfanion/workflow
Maintainers
Readme
create-opencode-workflow
⚡ This is an alias for @comfanion/workflow - same package, easier to remember!
AI-assisted development workflow with semantic code search, agents, and structured documentation.
Features
- 🔍 Semantic Code Search - Find code by meaning, not just text (
"authentication logic"→ finds auth handlers) - 🤖 AI Agents - Specialized personas (Analyst, PM, Architect, Developer) with skills
- 📝 Structured Workflow - PRD → Architecture → Epics → Stories → Implementation
- 🔄 Auto-indexing - Background indexing on startup with fun toast notifications
- 🎯 Jira Integration - Bidirectional sync with your project
Agents & Workflow
The workflow uses specialized AI agents, each with a unique persona and skills:
| Agent | Name | Role | Phase | |-------|------|------|-------| | 📊 Analyst | Sara | Requirements gathering, stakeholder interviews | Planning | | 📋 PM | Dima | PRD, epics, stories, sprint planning, Jira | Planning → Sprint | | 🏗️ Architect | Winston | System design, ADRs, coding standards | Planning | | 💻 Dev | Rick | TDD implementation, code review | Implementation | | ⚡ Coder | Morty | Quick implementation, bug fixes | Implementation | | 🔍 Researcher | Kristina | Technical/market/domain research | Any | | 🔄 Change Manager | Bruce | Documentation changes, impact analysis | Any |
Workflow Pipeline
Planning: /requirements → /prd → /coding-standards → /architecture
Sprint: /epics → /stories → /sprint-plan → /jira-sync
Development: /dev-story ↔ /code-review (loop until done)Key Skills
- requirements-gathering - Extract FR/NFR through interviews
- prd-writing - Product requirements documents
- architecture-design - Hexagonal/DDD patterns
- story-writing - User stories with Given/When/Then AC
- dev-story - Red-green-refactor implementation cycle
- jira-integration - Bidirectional sync with Jira
Quick Start
npx create-opencode-workflow initSemantic Code Search
Search your codebase by meaning, not just text matching:
# Terminal CLI:
npx create-opencode-workflow search "authentication logic"
npx create-opencode-workflow search "how to deploy" --index docs
npx create-opencode-workflow search "database config" --all
# AI agents use MCP tool automatically:
# search({ query: "authentication", index: "code" })How It Works
- Vectorizer converts code into embeddings using local AI model
- Indexes are stored in
.opencode/vectors/(code, docs, config) - Search finds semantically similar code chunks
- Auto-indexer keeps indexes fresh on startup
Available Indexes
| Index | Files | Use Case |
|-------|-------|----------|
| code | *.js, *.ts, *.py, *.go... | Find functions, classes, logic |
| docs | *.md, *.txt | Find documentation, guides |
| config | *.yaml, *.json | Find configuration, settings |
Commands
# Manual indexing
npx create-opencode-workflow index # Index all
npx create-opencode-workflow index --code # Index code only
npx create-opencode-workflow index --docs # Index docs only
# Check index status
npx create-opencode-workflow index --statusInstallation
NPX (recommended)
npx create-opencode-workflow initGlobal Install
npm install -g create-opencode-workflow
create-opencode-workflow initAlternative Package Name
npx @comfanion/workflow initCommands
init
Initialize .opencode/ in current project.
npx create-opencode-workflow initInteractive prompts:
- Your name - For personalized agent communication
- Communication language - Ukrainian, English, Russian
- Development methodology - TDD or STUB
- Vectorizer - Enable semantic search
- Jira integration - Enable/disable
Flags:
| Flag | Description |
|------|-------------|
| -y, --yes | Skip prompts, use defaults |
| --tdd | Use TDD methodology |
| --stub | Use STUB methodology |
| --jira | Enable Jira integration |
| --full | Create full repository structure |
update
Update .opencode/ to latest version.
npx create-opencode-workflow updatePreserves:
- ✅ Your
config.yaml(with comments!) - ✅ Vector indexes (
.opencode/vectors/) - ✅ Custom settings
doctor
Check installation health.
npx create-opencode-workflow doctorvectorizer
Manage semantic search vectorizer.
npx create-opencode-workflow vectorizer install # Install dependencies
npx create-opencode-workflow vectorizer status # Check statusConfiguration
config.yaml
# User settings
user_name: "Developer"
communication_language: "en" # en, uk, ru
# Development
development:
methodology: tdd # tdd or stub
# Semantic Search
vectorizer:
enabled: true
auto_index: true # Auto-index on startup
debounce_ms: 5000
indexes:
code: { enabled: true }
docs: { enabled: true }
config: { enabled: false }
exclude:
- "node_modules/**"
- "dist/**"
- "*.min.js"
# Jira Integration
jira:
enabled: false
url: "https://your-domain.atlassian.net"
project_key: "PROJ"What Gets Created
.opencode/
.opencode/
├── config.yaml # Your configuration
├── FLOW.yaml # Workflow definition
├── agents/ # AI agent personas
│ ├── analyst.md # Business Analyst
│ ├── pm.md # Product Manager
│ ├── architect.md # Solution Architect
│ └── dev.md # Senior Developer
├── skills/ # Knowledge modules (25+)
├── plugins/ # Auto-indexer plugin
├── vectorizer/ # Semantic search engine
│ ├── index.js
│ └── package.json
├── vectors/ # Vector indexes (auto-created)
│ ├── code/
│ ├── docs/
│ └── config/
├── tools/ # MCP tools
│ ├── search.ts # Semantic search tool
│ └── codeindex.ts # Index management tool
└── commands/ # Slash commandsdocs/
docs/
├── sprint-artifacts/ # Epics, stories, sprints
├── requirements/ # Requirements documents
├── architecture/ # Architecture + ADRs
└── coding-standards/ # Coding patternsAuto-Indexer Plugin
The auto-indexer runs on Claude Code / AI assistant startup:
- 🔍 Checks if indexes need updating
- 📊 Shows toast notification with file count
- ☕ Shows fun message while indexing ("Grab a coffee!")
- 📝 Logs to
.opencode/indexer.log
Disable auto-indexing:
# config.yaml
vectorizer:
auto_index: falseMethodologies
TDD (Test-Driven Development)
1. Write failing test (RED)
2. Write minimal code to pass (GREEN)
3. Refactor (BLUE)
4. RepeatSTUB (Stub-First Development)
1. Write interface/stub with TODO
2. Write tests against stub
3. Implement stub
4. Remove TODOsJira Integration
Set credentials:
export JIRA_EMAIL="[email protected]"
export JIRA_API_TOKEN="your-api-token"Requirements
- Node.js >= 18
- ~100MB disk for vectorizer dependencies
Links
- npm: https://www.npmjs.com/package/create-opencode-workflow
- Main package: https://www.npmjs.com/package/@comfanion/workflow
- GitLab: https://gitlab.com/comfanion/workflow
Inspired By
- BMAD Method - Breakthrough Method of Agile AI-Driven Development
- OpenCode - AI-native code editor
License
MIT
