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

Install from GitHub (recommended)

In Claude Code, run:

/plugin

1. Navigate to the Marketplaces tab
2. Select Add marketplace and enter: masterleopold/anty-framework
3. Navigate to the Discover tab
4. Find anty and install it

After this, just run claude — Anty loads automatically in every session. No extra flags needed.

You can choose a scope during installation:

Try It (one-time)

git clone https://github.com/masterleopold/anty-framework.git
claude --plugin-dir ./anty-framework

This loads the plugin for the current session only. You'll need to pass --plugin-dir each time.

Updating

/plugin

Navigate to the Installed tab and select Update next to anty, or run /reload-plugins to refresh all plugins.

Requirements

• 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

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:

Subagents

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.

Design Theory

Anty's theoretical backbone — the five-axis model of autonomous workforce AI (Persistent Intent, Cheap Continuous Sensing, Expensive Re-Planning, Agent-to-Agent Protocol, Non-Monotonic Learning) and the five core design principles (Anticipative Execution, Progressive Trust, Strategy Kernel, Cost-Tiered Cognition, Persistent Identity Files) — is maintained as proprietary documentation in the private anty web-app repository. This plugin is a partial implementation of those principles.

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