How to Create Architecture
Use the architecture workflow to make technical decisions explicit and prevent agent conflicts during implementation.
When to Use This
Section titled “When to Use This”- Multi-epic projects (BMad Method, Enterprise)
- Cross-cutting technical concerns
- Multiple agents implementing different parts
- Integration complexity exists
- Technology choices need alignment
When to Skip This
Section titled “When to Skip This”- Quick Flow (simple changes)
- BMad Method Simple with straightforward tech stack
- Single epic with clear technical approach
Prerequisites
Section titled “Prerequisites”- BMad Method installed
- Architect agent available
- PRD completed
1. Load the Architect Agent
Section titled “1. Load the Architect Agent”Start a fresh chat and load the Architect agent.
2. Run the Architecture Workflow
Section titled “2. Run the Architecture Workflow”*create-architecture3. Engage in Discovery
Section titled “3. Engage in Discovery”This is NOT a template filler. The architecture workflow:
- Discovers technical needs through conversation
- Proposes architectural options with trade-offs
- Documents decisions that prevent agent conflicts
- Focuses on decision points, not exhaustive documentation
4. Document Key Decisions
Section titled “4. Document Key Decisions”Work with the agent to create Architecture Decision Records (ADRs) for significant choices.
5. Review the Architecture
Section titled “5. Review the Architecture”The agent produces a decision-focused architecture document.
What You Get
Section titled “What You Get”An architecture.md document containing:
- Architecture Overview - System context, principles, style
- System Architecture - High-level diagram, component interactions
- Data Architecture - Database design, state management, caching
- API Architecture - API style (REST/GraphQL/gRPC), auth, versioning
- Frontend Architecture - Framework, state management, components
- Integration Architecture - Third-party integrations, messaging
- Security Architecture - Auth/authorization, data protection
- Deployment Architecture - CI/CD, environments, monitoring
- ADRs - Key decisions with context, options, rationale
- FR/NFR-Specific Guidance - Technical approach per requirement
- Standards and Conventions - Directory structure, naming, testing
ADR Format
Section titled “ADR Format”## ADR-001: Use GraphQL for All APIs
**Status:** Accepted | **Date:** 2025-11-02
**Context:** PRD requires flexible querying across multiple epics
**Decision:** Use GraphQL for all client-server communication
**Options Considered:**1. REST - Familiar but requires multiple endpoints2. GraphQL - Flexible querying, learning curve3. gRPC - High performance, poor browser support
**Rationale:**- PRD requires flexible data fetching (Epic 1, 3)- Mobile app needs bandwidth optimization (Epic 2)- Team has GraphQL experience
**Consequences:**- Positive: Flexible querying, reduced versioning- Negative: Caching complexity, N+1 query risk- Mitigation: Use DataLoader for batchingExample
Section titled “Example”E-commerce platform produces:
- Monolith + PostgreSQL + Redis + Next.js + GraphQL
- ADRs explaining each choice
- FR/NFR-specific implementation guidance
- Focus on decisions that prevent agent conflicts
- Use ADRs for every significant technology choice
- Keep it practical - don’t over-architect
- Architecture is living - update as you learn
Next Steps
Section titled “Next Steps”After architecture:
- Create Epics and Stories - Work breakdown informed by architecture
- Implementation Readiness - Gate check before Phase 4
Related
Section titled “Related”- Create PRD - Requirements before architecture
- Create Epics and Stories - Next step
- Run Implementation Readiness - Gate check
- Why Solutioning Matters