OpenCode BMAD Workflow Plugin: AI-Driven Full-Process Automation for Software R&D

This plugin integrates the BMAD methodology into OpenCode, using 5 professional AI agents to automate R&D processes like product planning, requirement analysis, architecture design, iteration management, and code review. It supports both fully automatic and interactive modes, model customization, and structured document management, aiming to improve the collaboration efficiency of software teams.

Traditional R&D processes rely heavily on manual coordination and document writing, which are time-consuming and labor-intensive. BMAD (Business Model and Architecture Development) is a structured R&D method that emphasizes systematic progression from business models to architecture design. The OpenCode BMAD plugin combines BMAD with AI agents, built for OpenCode (an open and extensible AI programming tool), to address the pain points of traditional R&D.

1. Professional Agents: Each agent has a clear role; 2. Dual Mode Support: Fully automatic mode (using workflow_* commands) requires no manual intervention, while interactive mode (using /workflow_* commands) needs user confirmation at key steps; 3. Structured Output: Documents are stored in the .workflow/ directory for easy management; 4. Model Agnosticism: Supports LLM providers like Anthropic, OpenAI, Ollama, etc., requiring tool calling capabilities.

The plugin implements 5 core workflows: 1. Product Roadmap (generated by the Analyst Agent); 2. Epic Definition (detailed documents and feature lists written by the PM Agent); 3. Feature Implementation (collaboration between PM, Architect, and PM Agents to complete PRD, architecture design, and task breakdown); 4. Iteration Planning (PM Agent generates user stories with acceptance criteria and effort estimation); 5. Code Review (Analyst + Reviewer Agents complete analysis and formal reports).

• Agent Configuration: 6 types of agents (PM, Architect, Analyst, etc.), with default models like qwen3-coder:30b for PM and gemma4:e4b for Architect, supporting customization; - Installation Steps: Copy files to OpenCode's configuration directory, install dependencies, register the plugin, restart OpenCode; - File Structure: Includes directories like agents/ (agent definitions), commands/ (interactive commands), plugin/ (core logic), etc.

Value: Standardizes processes, automates repetitive tasks to improve efficiency, accumulates knowledge, and ensures quality through multi-agent collaboration; Limitations: Relies on LLM capabilities, context window constraints, and requires manual review; Future Directions: Smarter task breakdown, CI/CD integration, support for more language frameworks, data-driven optimization suggestions.