Skip to content

Adanmohh/organtic

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

49 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Organtic

The AI-native organization. 25 experts from idea to launch.

Organtic is a team of 25 AI experts that plug into Claude Code. Run them individually or as coordinated Agent Teams — multiple Claude Code instances working in parallel with shared task lists, inter-agent messaging, and a team lead that synthesizes results. Each expert applies named frameworks. Not generic advice.

The name: organic + agentic. An organization that grows with your project.

The Organization

                              ORGANTIC
       ┌──────────────────────────────────────────────────────┐
       │                   EXPLORE TEAM                       │
       │                                                      │
       │  vision (1)  ──→  validate (3)  ──→  build (5)      │
       │  What are we      Is this worth      What are we     │
       │  doing?           building?          building?       │
       │                                                      │
       │               ──→  launch (7)  ──→  craft (4)       │
       │                    How do we         Build it         │
       │                    bring it to                        │
       │                    market?                            │
       ├──────────────────────────────────────────────────────┤
       │                  EXPLOIT TEAM                        │
       │                                                      │
       │  document (2)    present (1)    studio (3)           │
       │  Write it up     Show it        Film it              │
       └──────────────────────────────────────────────────────┘

       ORCHESTRATION
       pipeline — Chains all experts with Brief enforcement,
                  conditional routing, and organizational memory

How It Works

Install Organtic plugins into Claude Code. Each plugin adds a team of domain experts. Run them three ways:

  1. Individual expert/value-mapper My idea runs one expert
  2. Agent Team/validate:full-team My idea spawns 3 teammates as separate Claude Code instances that coordinate via shared task lists and message each other
  3. Pipeline/pipeline:run My idea chains all experts across 7 phases with a constitutional Brief, cross-session memory, and adaptive routing

Agent Teams use Claude Code's experimental Agent Teams feature. Each teammate is a full Claude Code session with its own context window. Unlike subagents (which just report results back), teammates share findings, challenge each other's assumptions, and coordinate independently.

Every expert applies specific frameworks (Osterwalder VPC, Hormozi Value Equation, Reforge Growth Loops — not vague "best practices"), and produces structured deliverables.

Quick Start

# Install from your terminal
npx organtic

# Or add the marketplace manually in Claude Code
/plugin marketplace add Adanmohh/organtic
/plugins   # select which teams to install

Enable Agent Teams (required for :full-team commands):

// ~/.claude/settings.json
{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  }
}

Run your first expert:

/value-mapper AI tool that helps nurses track patient vitals

Run a full Agent Team:

/validate:full-team BurnoutLab - AI-powered burnout recovery courses

Run the full pipeline:

/pipeline:run BurnoutLab - AI-powered burnout recovery courses

Architecture: Agentic Organization

Organtic mirrors a human organization — each plugin is a department, each expert is a specialist, and the Brief is the company charter.

┌─────────────────────────────────────────────────────┐
│                    THE BRIEF                         │
│         Constitutional document created in           │
│         Phase 0 (Vision Anchor). Defines             │
│         non-negotiables, vision, deliverables.       │
│         No expert can override it.                   │
└──────────────┬──────────────────────┬───────────────┘
               │                      │
    ┌──────────▼──────────┐  ┌───────▼────────────┐
    │   SubagentStart     │  │   SubagentStop     │
    │   Auto-injects      │  │   Semantic QA      │
    │   Brief into every  │  │   checks output    │
    │   spawned expert    │  │   against Brief    │
    └─────────────────────┘  └────────────────────┘
               │                      │
    ┌──────────▼──────────────────────▼───────────┐
    │              Stop Hook                       │
    │   Phase orchestrator: advances phases,       │
    │   conditional routing, error recovery,       │
    │   execution mode enforcement                 │
    └──────────────────────────────────────────────┘

Three Enforcement Layers

  1. SubagentStart hook — Every spawned expert automatically receives the Brief's non-negotiables, vision, and core deliverables. No expert starts work without the company charter.

  2. SubagentStop hook — Dual-layer QA:

    • Command hook: Fast programmatic checks (language, deliverable keywords)
    • Prompt hook: Semantic evaluation — an LLM checks if the expert's output actually aligns with Brief requirements, not just keyword matching
  3. Stop hook — Phase orchestrator that:

    • Advances through phases with full context handoff
    • Routes dynamically based on phase output signals
    • Retries failed phases (max 2 attempts) with graceful degradation
    • Injects execution mode instructions (sequential/parallel/teams)

Organizational Memory

Memory Type What It Stores Persists Across
Brief Non-negotiables, vision, deliverables All phases in a run
Phase summaries Each department's findings Later phases in same run
Experiential learnings What worked/failed across projects Projects
Founder preferences Communication style, risk tolerance Sessions

Agent Teams vs Subagents vs Pipeline

Individual Expert Agent Team (:full-team) Pipeline
Sessions 1 Multiple (1 lead + N teammates) 1 per phase (sequential)
Communication None Teammates message each other Phase outputs saved to memory
Coordination None Shared task list with dependencies Stop-hook advances phases
Brief enforcement None None Full (inject + QA + routing)
Best for Quick single-domain analysis Deep multi-expert collaboration Full end-to-end engagement
Token cost Low High (each teammate = separate instance) Medium (sequential phases)

The Pipeline

The pipeline orchestrates all experts across 7 phases. Each phase saves findings to cross-session memory. Later phases build on earlier work. The Brief (created in Phase 0) acts as the constitutional document that every expert must respect.

Phase 0: vision              1 step (Vision Anchor)
          └─ Creates the Brief — non-negotiables, vision,
             deliverables, business model, constraints

Phase 1: validate            3 experts
          ├─ Value Mapper
          ├─ Business Modeler
          └─ Experiment Designer

Phase 2: build               5 experts
          ├─ Product Strategist
          ├─ Brand Strategist
          ├─ Growth Designer
          ├─ Spec Writer
          └─ Product Designer

Phase 3: launch-strategy     3 experts
          ├─ Offer Designer
          ├─ Copywriter
          └─ Funnel Architect

Phase 4: launch-coherence    1 expert (audits all strategy outputs)
          └─ Growth Auditor

Phase 5: launch-execution    3 experts
          ├─ Media Planner
          ├─ SEO Strategist
          └─ AI Search Optimizer

Phase 6: craft               4 experts
          ├─ Landing Page Builder
          ├─ MVP Scaffolder
          ├─ Component Designer
          └─ Tester

Conditional Routing

Phases can signal the pipeline to skip or loop back based on findings:

vision → validate ──→ build ──→ launch → craft
              ↑          │         │
              └──────────┘         │
              └────────────────────┘

Experts include routing signals in their output:

  • [ROUTE:skip-build] — consulting offer, no product to build
  • [ROUTE:loop-back-to-validate] — assumptions need re-validation
  • [ROUTE:loop-back-to-vision] — fundamental pivot needed
  • [ROUTE:skip-to-craft] — skip launch, go straight to building

Loop protection caps re-visits at 2 per phase.

Error Recovery

  • Failed phases retry up to 2 times with detailed instructions
  • After max retries, the phase is skipped with a warning injected into the next phase
  • Brief violations get 3 attempts before being accepted
  • Phase health is tracked in the pipeline state file

Pipeline Options

# Full pipeline (starts from vision)
/pipeline:run MyProject - description of the idea

# Start from a specific phase (builds on prior memory)
/pipeline:run MyProject --start-phase build

# Autonomous mode: experts research before asking questions
/pipeline:run MyProject --autonomous

# Parallel mode: dispatch experts as concurrent sub-agents
/pipeline:run MyProject --parallel

# Agent Teams mode: experts coordinate as teammates (experimental)
/pipeline:run MyProject --teams

# Combine flags
/pipeline:run MyProject --autonomous --parallel

# Read idea from a brief file
/pipeline:run BurnoutLab.md

# Override an active pipeline
/pipeline:run MyProject --force

# Pipeline management
/pipeline:status    # View current state + activity log
/pipeline:cancel    # Stop the active pipeline
/pipeline:resume    # Resume from any phase

Execution Modes

Mode Flag How experts run
Sequential (default) One expert at a time, findings passed forward
Parallel --parallel All phase experts dispatched as concurrent sub-agents
Teams --teams Experts coordinate as Agent Team members (experimental)

Autonomous Mode

With --autonomous, every expert follows the 80/20 Research Protocol before analysis:

  1. Broad Scan — 5-8 web searches per expert across competitors, market data, pricing, and customer pain points
  2. Deep-Dive — 3-5 follow-up queries on the biggest unknowns
  3. Synthesize — Apply frameworks using research data, not assumptions
  4. Ask only the unanswerable — founder intent, budget constraints, strategic preferences

Sub-agents run searches in parallel across different research domains. The result: experts that come to you with findings instead of questions.

Pipeline Observability

Every hook writes structured events to .claude/pipeline-log.jsonl:

{"timestamp":"2026-03-08T14:30:00Z","event":"phase_complete","phase":"validate","message":"3 experts completed"}
{"timestamp":"2026-03-08T14:31:00Z","event":"brief_injected","phase":"build","message":"agent_type: general-purpose"}
{"timestamp":"2026-03-08T14:35:00Z","event":"qa_check_passed","phase":"build","message":"Expert output OK"}

View the log with /pipeline:status.

Explore Team (15 experts)

validate — Is this worth building?

Expert Frameworks
Value Mapper Osterwalder VPC + Jobs-to-be-Done + Identity Jobs
Business Modeler BMC + AI-BMC + Blue Ocean Strategy + Pricing Research
Experiment Designer Strategyzer Testing + ICE/RICE + AI Experiments

build — What are we building?

Expert Frameworks
Product Strategist Reforge + Growth Loops + Cagan's Transformed
Brand Strategist Marty Neumeier Brand Gap + Design Tokens
Growth Designer Torres AI Discovery + Hooked Model + PLG
Spec Writer Shape Up methodology + AI Feature Shaping
Product Designer Don Norman + JTBD UX + AI UX Patterns

launch — How do we bring it to market?

Expert Frameworks
Offer Designer Hormozi Value Equation + Grand Slam Offer + $100M Money Models
Copywriter Copyhackers + Ogilvy + Gary Vee Content Systems
Funnel Architect Brunson Value Ladder + Suby 8-Phase + AI-native Funnels
Growth Auditor Reforge Racecar + Retention Curves + Channel-Model Fit
Media Planner Dennis Yu Dollar-a-Day + 3x3 Grid + Meta Advantage+
SEO Strategist CXL Technical SEO + Search Intent Mapping + GEO
AI Search Optimizer CXL Entity SEO + LLM Visibility Optimization

Exploit Team (10 agents)

craft — Build it

Agent What it produces
Landing Page Builder Conversion-optimized landing pages (React/HTML)
MVP Scaffolder Full-stack prototypes with auth, DB, and API
Component Designer Polished UI components (React + Tailwind)
Tester Strategy-informed persona-based QA with Playwright

document — Write it up

Agent What it produces
Business Writer Business plans, SOPs, case studies, one-pagers
Pitch Deck Writer Investor pitch decks and sales proposals (Marp to PDF/PPTX)

present — Show it

Agent What it produces
Slide Designer Professional slide decks from strategy outputs (Marp markdown)

studio — Film it

Agent What it produces
Demo Producer Full product demo videos (Playwright + ElevenLabs + Remotion)
Clip Maker Short social clips for Instagram, TikTok, YouTube, LinkedIn
Content Creator Branded social media images and short-form videos

Usage

Run individual experts

/value-mapper AI tool that helps nurses track patient vitals
/offer-designer B2B SaaS at $99/mo for HR teams
/media-planner $3k/mo budget for Facebook and Google ads

Run an Agent Team

/validate:full-team My new SaaS idea for project management
/build:full-team Continue building on the validated idea
/launch:full-team Go to market with the product strategy

Each :full-team command spawns an Agent Team where experts run as separate Claude Code instances, coordinate through a shared task list, and message each other directly.

Use exploit agents

/landing-page Build a landing page from our launch outputs
/mvp Scaffold an MVP for our validated idea
/pitch-deck Create an investor pitch deck
/slides Turn our strategy into a board presentation
/product-demo Record a demo video of our web app
/social-clip Create Instagram Reels from our demo

Cross-Session Memory

Each phase outputs project-scoped tags that are auto-captured by claude-mem:

[VISION:anchor-complete:ProjectName]   → Brief + research findings
[VALIDATE:pipeline-complete:ProjectName]  → value prop + business model + experiments
[BUILD:pipeline-complete:ProjectName]     → product strategy + brand + specs + UX
[LAUNCH:strategy-complete:ProjectName]    → offer + copy + funnel
[LAUNCH:coherence-complete:ProjectName]   → growth model + priorities + risks
[LAUNCH:pipeline-complete:ProjectName]    → full go-to-market + execution plans
[CRAFT:pipeline-complete:ProjectName]     → built artifacts + test results

Later phases search memory by date range and project name to build on validated findings. No manual copy-pasting between sessions.

Organizational memory also stores experiential learnings (what worked across projects) and founder preferences (communication style, risk tolerance) that improve future pipeline runs.

Requirements

  • Claude Code CLI
  • Agent Teams enabled (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1) for :full-team commands
  • Optional: claude-mem for cross-session memory (enables pipeline phase handoffs and organizational learning)

For studio plugin

  • Node.js + npm
  • Playwright (npm install playwright)
  • Remotion (npx create-video@latest)
  • ffmpeg (scoop install ffmpeg)
  • ElevenLabs API key (for voiceover)

License

MIT

About

The AI-native organization. 25 experts from idea to launch. Agentic pipeline with constitutional Brief enforcement, conditional routing, and organizational memory. Claude Code plugin ecosystem.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors