What is BMAD?

BMAD stands for Breakthrough Method of Agile AI-driven Development. It is a structured approach to AI-assisted coding that separates planning from execution, ensuring the agent thinks before it acts.

Most chat-based coding agents operate in freeform mode - they receive a prompt and immediately start generating code. This works for simple tasks, but complex projects suffer from scope creep, missed requirements, and wasted tokens on wrong approaches.

BMAD solves this by enforcing three distinct phases: Plan, Execute, and Verify.

Planning Phase

When you give 8gent a task, it first classifies the work into one of five categories:

Each category has a tailored planning approach. Code tasks get numbered implementation steps. Research tasks get exploration avenues. Creative tasks get structured briefs.

How Planning Looks

User: Build a Next.js site with auth and dark mode

8gent:
┌─────────────────────────────────────────┐
│ PLAN                                    │
├─────────────────────────────────────────┤
│ 1. Initialize Next.js project           │
│ 2. Set up authentication (NextAuth)     │
│ 3. Create theme provider (dark/light)   │
│ 4. Build UI components                  │
│ 5. Add tests and verify                 │
└─────────────────────────────────────────┘

The plan is visible at any time via /plan. It also populates the Kanban board automatically - each step becomes a card that moves from Backlog to In Progress to Done as 8gent works.

Execution Phase

Once the plan is set, 8gent executes each step sequentially. During execution, the system tracks:

• Modified files - Every file written or edited
• Tool calls - Every command run, search performed, or file read
• Errors encountered - Build failures, type errors, test failures
• Evidence collected - Pass/fail results from verification checks

The proactive planner updates predictions after every tool call, adjusting the remaining steps based on what it learns.

Foresight and Avenues

8gent does not just follow a linear plan. It anticipates problems and explores alternatives:

• Prediction - Analyzes the current task to predict the next 3-5 logical steps. View with /predict or /next.
• Avenues - Before committing to an approach, 8gent briefly explores multiple options and weighs trade-offs. View with /avenues.
• Proactive warnings - If 8gent detects scope creep, technical debt, circular dependencies, over-engineering, or performance traps, it speaks up.

Warning: Adding a 4th boolean prop to <Button> suggests
it may be time for a variant-based API instead.

Shall I propose a refactor, or continue as planned?

Momentum Tracking

The planner tracks execution velocity:

• Steps completed - How many plan steps are done
• Rate - Steps per minute
• Streak - Consecutive successful steps

View momentum stats with /momentum.

Verification Phase

After execution, 8gent collects evidence that the work is correct:

• Type checking - Runs tsc to verify TypeScript compiles
• Tests - Runs test suites if they exist
• Lint - Checks for code quality issues
• Build - Verifies the project builds successfully
• Manual checks - Reads back modified files to confirm changes

Evidence badges (pass/fail) appear in the chat stream after each file write and command execution. At the end of a task, 8gent provides a completion report summarizing all evidence.

View the full evidence breakdown with /evidence.

How BMAD Differs from Freeform Agents

BMAD Documents

8gent ships with 20 BMAD planning documents covering project briefs, product requirement documents (PRDs), architecture docs, and epics. These live in docs/bmad/ within the 8gent-code repository and serve as templates for the agent's planning process.