LaunchKit v2

A multi-agent marketing strategy platform powered by Claude Code. 11 specialized AI agents collaborate to produce a complete, professional marketing strategy — from client discovery through executive summary.

What's New in v2

• 11 specialized agents instead of one monolithic prompt
• 10 phases (up from 6) — added SEO, Analytics, and Executive Summary
• Parallel execution — Phases 4, 5, 6 run simultaneously
• QA reviewer cross-checks every deliverable before you see it
• Local web dashboard for intake, progress tracking, and output viewing

Quick Start

Option A: CLI Only (Recommended to start)

cd launch-kit
claude

Claude will ask about your client and run the full pipeline. Use slash commands for control:

• /full-pipeline — Run all 10 phases end-to-end
• /run-phase 3 — Run a specific phase
• /status — See pipeline progress

Option B: CLI + Web Dashboard

# Terminal 1: Start the dashboard
cd launch-kit/web
npm install
npm start

# Terminal 2: Run Claude
cd launch-kit
claude

The web UI opens at http://localhost:3000. You can:

1. Fill out the client intake form in the browser
2. Watch phases complete in real-time
3. Review and approve outputs
4. Request changes (creates feedback files Claude picks up)
5. Export deliverables as markdown or JSON

Option C: Provide a Brief First

Drop any client info into the inputs/ folder — a brief, brand guidelines PDF, competitor URLs, notes, whatever you have. Then run claude.

Architecture

launch-kit/
├── CLAUDE.md                  ← Orchestrator brain
├── .claude/
│   ├── agents/                ← 11 specialized subagents
│   │   ├── intake.md
│   │   ├── researcher.md
│   │   ├── brand-strategist.md
│   │   ├── funnel-architect.md
│   │   ├── ad-strategist.md
│   │   ├── outreach-specialist.md
│   │   ├── seo-strategist.md
│   │   ├── analytics-architect.md
│   │   ├── content-creator.md
│   │   ├── executive-summarizer.md
│   │   └── quality-reviewer.md
│   ├── commands/              ← Slash commands
│   │   ├── run-phase.md
│   │   ├── full-pipeline.md
│   │   └── status.md
│   └── settings.local.json   ← Tool permissions
├── inputs/                    ← Client briefs + web UI intake
├── outputs/                   ← Generated deliverables + progress.json
├── templates/                 ← Output scaffolds (agents read these)
├── web/                       ← React dashboard (optional)
├── .mcp.json                  ← MCP server config
└── README.md                  ← You're here

Phase Pipeline

0 (Intake) → 1 (Research) → 2 (Brand) → 3 (Funnel)
                                            │
                              ┌──────────────┼──────────────┐
                              ▼              ▼              ▼
                         4 (Ads)      5 (Outreach)     6 (SEO)
                              │              │              │
                              └──────────────┼──────────────┘
                                             ▼
                                       7 (Analytics)
                                             ▼
                                       8 (Content)
                                             ▼
                                    9 (Executive Summary)

Phases 4, 5, and 6 run in parallel after Phase 3 completes.

What You Get

After running the full pipeline, your outputs/ folder contains:

File	Phase	What's In It
00-client-profile.md	0	Compiled client profile
01-deep-research.md	1	Company, competitor, market, and audience research
02-brand-voice.md	2	Brand personality, voice, messaging pillars, positioning
03-funnel-strategy.md	3	Customer journey and funnel design
04-ad-strategy.md	4	Paid channels, targeting, budget, creative direction
05-outreach-plan.md	5	Email sequences, DM templates, partnerships
06-seo-strategy.md	6	Keywords, content pillars, technical SEO
07-analytics-plan.md	7	Tracking, dashboards, attribution, reporting
08-content-pack.md	8	Social posts, headlines, CTAs, emails, landing page copy
09-executive-summary.md	9	Strategic overview + 30-day action plan
progress.json	—	Pipeline state (web UI reads this)

Customizing

• Edit CLAUDE.md to change the orchestration flow, add/remove phases, or modify the routing table
• Edit agent files in .claude/agents/ to customize individual specialist behavior
• Edit templates in templates/ to change the output format for each deliverable
• Edit slash commands in .claude/commands/ to add custom workflows

How It Works

1. Orchestrator (CLAUDE.md) reads inputs and delegates to specialized agents
2. Each agent reads its dependencies, reads the template, writes its output
3. Quality reviewer checks every output for completeness, consistency, and accuracy
4. User approves or requests changes before the next phase starts
5. Progress.json tracks state — the web UI reads this for real-time updates
6. Web UI communicates with Claude via the filesystem — no API server needed

Requirements

• Claude Code CLI installed
• Node.js 18+ (for the web dashboard, optional)