pongogo 0.3.0

AI agent knowledge routing system for software development

Pongogo

Portable AI agent knowledge routing system. Install Pongogo on any repository to get intelligent instruction routing for AI coding assistants.

Quick Start

# Install Pongogo CLI
pip install pongogo

# Configure MCP server for Claude Code (requires Docker)
pongogo setup-mcp

# Initialize in your repository
cd your-project
pongogo init

This creates a .pongogo/ directory with configuration and seeded instruction files that help AI assistants understand your project's patterns and practices.

Installation

Zero-Config Install (Recommended)

The fastest way to get started:

curl -sSL https://get.pongogo.com | bash

The installer will:

1. Check if Docker is available
2. If Docker present: Pull the Docker image and configure Claude Code
3. If no Docker: Help you install Docker first

Docker Installation

Docker is required for the MCP server to ensure proper multi-repo isolation:

# Pull the image
docker pull ghcr.io/pongogo/pongogo:stable

# Configure Claude Code
pongogo setup-mcp

Why Docker? When using Pongogo across multiple repositories on the same machine, Docker ensures each workspace gets isolated state via volume mounts.

Requirements

• Docker (required for MCP server)
• Python 3.10+ (for CLI tools)
• Claude Code installed

Usage

Initialize Pongogo

pongogo init

Creates a .pongogo/ directory with:

• config.yaml - Configuration for enabling/disabling instruction categories
• instructions/ - Seeded instruction files (42 files across 14 categories)

Command Options

Flag	Short	Description
--minimal	-m	Install only core instruction categories
--force	-f	Overwrite existing .pongogo/ directory
--no-interactive	-y	Accept all defaults without prompting

Configuration

After initialization, customize .pongogo/config.yaml:

# Enable/disable instruction categories
categories:
  software_engineering: true
  project_management: true
  agentic_workflows: true
  # ... set to false to disable

# Customize placeholders for your project
placeholders:
  wiki_path: wiki/
  docs_path: docs/
  owner_repo: your-org/your-repo

Instruction Categories

Category	Files	Description
software_engineering	3	Git safety, commit formats, Python standards
project_management	6	Work logging, scope prevention, task management
agentic_workflows	4	Agent decision making, compliance patterns
safety_prevention	3	Validation-first execution, systematic prevention
trust_execution	3	Trust-based execution, feature development
_pongogo_core	10	Core Pongogo workflows
+ 8 more categories

MCP Server Integration

Pongogo includes an MCP (Model Context Protocol) server that routes instructions to AI coding assistants.

Configure Claude Code

# Configure Claude Code (requires Docker)
pongogo setup-mcp

# Preview configuration without changes
pongogo setup-mcp --dry-run

Upgrading

# Check for updates
/pongogo-status

# Upgrade
/pongogo-upgrade
# or: docker pull ghcr.io/pongogo/pongogo:stable

Architecture

Pongogo routes instructions based on context:

User message → MCP Server → Semantic Router → Matched Instructions → Agent

The routing engine uses:

• Pattern matching: Keyword and regex patterns
• Context disambiguation: Positive/negative pattern weights
• Lexicon database: 329 entries for guidance, friction, and hedging detection

Status

• ✅ pongogo init CLI
• ✅ Seeded instruction files (42 files, 14 categories)
• ✅ MCP server with Docker distribution
• ✅ Claude Code integration
• ✅ CI/CD with release train (alpha/beta/stable)

License

MIT License - see LICENSE

---

Named after "Pongo" — the scientific name for the orangutan genus — reflecting the intelligent, collaborative nature of the system.