BMAD Planning & Orchestrator

A Claude Code plugin that plans and orchestrates. It never writes the code.

Attribution

What this plugin is

Modern coding harnesses are good at writing code. They are much weaker at the part that actually determines whether parallel AI development succeeds: planning the work so that many agents can build at once without colliding.

That is exactly what the BMAD Method is good at. This plugin focuses BMAD squarely on its strongest contribution — upstream planning and orchestration — and deliberately stops at the handoff.

It produces the artifacts that let other tools do the dev work fast, in parallel, and without conflicts:

• One architecture so every agent shares API style, data model, naming, and security approach — preventing semantic conflicts that turn parallel work into an integration nightmare. (Catching alignment in solutioning is roughly 10x cheaper than discovering it mid-build.)
• Stories scoped to disjoint files, dependency-ordered, grouped into parallel waves — preventing file/merge conflicts that serialize teams.
• A tool-agnostic handoff-manifest.json any dev runner can consume.

What it does NOT do

No code generation. No running tests, linting, or coverage checks. No reviewing implemented diffs. The last artifact this plugin produces is a ready-for-dev story file and the handoff manifest. Implementation is your dev tool’s job.

Install

From the Claude Code plugin marketplace:

/plugin marketplace add aj-geddes/claude-code-bmad-skills
/plugin install bmad-planning-orchestrator@bmad-method-harness
/reload-plugins

Skills are namespaced /bmad-planning-orchestrator:<skill> and most auto-invoke based on what you are doing.

For local development and testing:

claude --plugin-dir ./bmad-planning-orchestrator

The planning thesis

This plugin exists because planning is the multiplier. When the upstream work is done well — a shared architecture, stories scoped to owned files, a dependency-ordered wave plan — an AI dev tool can fan out across many parallel workstreams with no coordination overhead. When the upstream work is skipped, parallel dev produces collisions, rework, and integration pain.

The BMAD Method provides exactly the planning discipline that makes parallel AI development reliable. This plugin harnesses it.

The flow

BMAD’s four phases, right-sized by an interactive track selection at init time:

bmad-init ──▶ select track, create workspace + decision-log + project-context
    │
ANALYSIS (optional)   brainstorm · research · product-brief · prfaq · spec
    │
PLANNING              prd  (or tech-spec for Quick Flow)
    │
SOLUTIONING           ux (if UI) · architecture · epics-and-stories · readiness-check
    │                                                   └── "Planning Ends Here"
ORCHESTRATION         sprint-planning · parallel-plan · handoff
    │
    ▼
ready-for-dev story files + handoff-manifest.json  ──▶  YOUR DEV TOOL

bmad-help is the spine: at any point it reads your artifact state and tells you which skill to run next. Sizing is one-dev-day story decomposition and count-based delivery — no Fibonacci points, no velocity, no burndown.

Skills overview

The plugin ships 20 skills across four groups:

Orchestration spine

Analysis (all optional — enter at any point)

Planning

Solutioning

Orchestration and handoff

Cross-phase and meta

Documentation

Community and support

• GitHub Issues: Report bugs or request features
• BMAD Method (upstream): github.com/bmad-code-org/BMAD-METHOD
• BMAD Discord: discord.gg/gk8jAdXWmj
• License: MIT