Anty Framework

Your AI co-worker for startup sales & marketing — powered by Claude Code.

A Claude Code plugin that gives startup founders a proactive AI co-worker for sales & marketing. Anty doesn't wait for instructions — it analyzes your situation, plans strategy, proposes actions, and executes with graduated autonomy.

You describe your business. Anty builds the go-to-market plan and runs it.

What is Anty?

Anty is an AI workforce agent that operates as your S&M (Sales & Marketing) team member. It combines 15+ strategic frameworks from business literature into an integrated planning and execution engine:

• Strategy Kernel (Rumelt) — Diagnosis → Guiding Policy → Coherent Actions
• Critical Chain (Goldratt/TOC) — Buffer management, relay runner execution, constraint focus
• QUEST (Wiss) — Socratic questioning methodology for deep understanding
• Nudge (Thaler) — Choice architecture for better decision-making
• Culture Map (Meyer) — 8-dimension cultural localization for global outreach
• Cold Start Problem (Chen) — Network effects, atomic networks, escape velocity
• Value Stick (Oberholzer-Gee) — WTP/WTS pricing strategy
• And more: Effectuation, Pre-Mortem, Plan B, Cash Machine pipeline, PMF detection...

Installation

# Clone the framework
git clone https://github.com/masterleopold/anty-framework.git

# Launch Claude Code with the plugin
claude --plugin-dir ./anty-framework

Or install permanently via marketplace:

# Add as a local marketplace
claude plugins marketplace add ./anty-framework

# Install the plugin
claude plugins install anty@anty-marketplace

Requires Claude Code CLI installed and authenticated.

Quick Start

/anty:onboard     # Conversational interview — builds your business context
/anty:plan         # Generates Strategy Kernel + KPI tree + action proposals
/anty:actions      # Review and approve 3-option action proposals
/anty:status       # CLI dashboard — goals, buffer health, pipeline
/anty:loop 30m     # Start persistent operation (scans every 30 min)

Commands

Command	Description
/anty:onboard	QUEST-based conversational interview (14 topics)
/anty:plan	Strategy Kernel + KPI tree + Pre-Mortem + scenarios
/anty:actions	3-option choice approval with Nudge architecture
/anty:status	CLI dashboard (goals, buffer, growth, pipeline)
/anty:pipeline	Sales pipeline management (Cash Machine)
/anty:scan	10-point analysis cycle
/anty:loop	Persistent operation at specified interval
/anty:rules	View/edit User Rules (highest priority)
/anty:identity	View/edit Agent Identity (communication style, values)
/anty:review	5-question review engine

How It Works

1. You onboard — /anty:onboard runs a 14-topic Socratic interview covering product, market, pricing, competitors, pain intensity, PMF signals, culture settings, and more
2. Anty plans — /anty:plan generates a Strategy Kernel (diagnosis with reframing, guiding policy, coherent actions), decomposes into a KPI tree, identifies the Crux (most important AND solvable challenge), runs pre-mortem analysis, and proposes 3-option actions
3. You choose — /anty:actions presents options with Nudge choice architecture (smart defaults, outcome mapping, your own data surfaced before choices)
4. Anty executes — Actions are executed by specialized subagents (Research, Planner, Content, Analyst) with graduated autonomy
5. Anty learns — Every choice you make trains the preference model. Over time, proposals get more accurate and autonomy increases
6. Anty scans — /anty:loop runs continuous analysis checking buffer health, KPI propagation, pipeline stalls, environmental changes, and anti-patterns

Domain Knowledge (24 Skills)

Anty's intelligence comes from 24 domain-specific knowledge modules across 7 domains:

Domain	Skills
Strategy	Strategy Kernel, Pre-Mortem & Plan B, Effectuation, Disruption Analysis
Planning	Critical Chain & TOC, KPI Tree, Root Cause Analysis, Review Engine
Behavioral	QUEST Interview, Choice Architecture, Confidence Calibration, Culture Map
Marketing	Growth Metrics & YC Benchmarks, Retention Cohorts & PMF, Network Effects, Content Rules
Sales	Pipeline (Cash Machine), Lead Qualification, Anticipatory Selling
Pricing	Value Stick (WTP/WTS), Unit Economics (CLV/CAC/NDR)
Failure Patterns	Anti-Pattern Detection (12 patterns), Failure Essence (4 patterns), Plan B Framework

Subagents

Agent	Role	Model
Research	Market research, competitor analysis, six-force environmental scanning	Opus
Planner	Strategy, KPI tree, crux identification, action generation	Opus
Content	Outreach drafts, content creation, cultural localization	Opus
Analyst	KPI tracking, verification, root cause analysis, anti-pattern detection	Opus

Data Storage

All data persists locally in .anty/ (gitignored — your data stays private):

.anty/
├── config.yaml              # Plugin config (culture, scan frequency, mode)
├── business-context.yaml    # Onboarding results
├── rules.md                 # User Rules (CLAUDE.md-style)
├── identity.md              # Agent Identity (SOUL.md-style)
├── scratchpad.md            # Working memory (append-only)
├── goals/                   # Goals with Strategy Kernels
├── drivers/                 # KPI tree drivers
├── actions/                 # Action items with status
├── pipeline/deals/          # Sales pipeline deals
├── learning/                # Choice history, preferences, templates
└── history/scans/           # Historical scan results

Key Design Principles

• Invisible AI — Frameworks are never front-loaded; they surface exactly when relevant (JIT knowledge delivery)
• Anti-sycophancy — Anty pushes back with evidence when your strategy conflicts with data
• Source attribution — "Your data shows..." never "I recommend..."
• Contextual confidence — Deflates overconfidence on irreversible decisions, protects momentum on reversible ones
• Cultural localization — Adapts communication to your culture (Layer A) and outreach to target market culture (Layer B)
• Customer as protagonist — All generated content puts the customer at center, never the company

Web App

Anty Framework is the Phase 0 prototype. The full web application (SaaS) is planned at 4n7y.com with:

• Real-time dashboard (Convex + Next.js)
• Data integrations (Stripe, PostHog, HubSpot via MCP)
• Cross-company learning (Action Performance Index)
• Stripe billing (per-action metering)

Plugin logic transfers directly to the web app — skills become knowledge modules, commands become dashboard screens, .anty/ files become database tables.

Prerequisites

• Claude Code installed and authenticated
• Claude Max or API subscription

Contributing

See CONTRIBUTING.md for guidelines on adding skills, commands, and agents.

License

MIT