peircean-abduction 1.2.2

A Logic Harness for abductive inference. Anomaly in → Hypothesis out.

Peircean Abduction

A Logic Harness for Abductive Inference.

"Abduction is the process of forming an explanatory hypothesis. It is the only logical operation which introduces any new idea." — Charles Sanders Peirce

Peircean Abduction is a Model Context Protocol (MCP) server that gives LLMs the ability to perform rigorous abductive reasoning. It forces models to generate multiple competing hypotheses for an observation and evaluate them using Inference to the Best Explanation (IBE).

✨ Features

• MCP-First Design: Built to be used directly within Claude Desktop, Cursor, or any MCP-compliant client.
• Provider Agnostic: Works with Anthropic, OpenAI, Gemini, and Ollama.
• No API Key Required (Default): By default, it generates prompts for you to run. You only need an API key if you want the tool to execute the reasoning loop autonomously.
• Council of Critics: Simulates a debate between a Logician, Empiricist, and Scientist to refine hypotheses.

🚀 Quick Start

1. Install

pip install peircean-abduction

2. Connect to Claude Desktop

Run the installer to automatically configure Claude Desktop:

peircean --install

3. Use it!

Open Claude and ask:

"Use the peircean_observe_anomaly tool to analyze why my server latency spiked but CPU usage is flat."

🛠️ Configuration

You can configure the provider and other settings via the interactive wizard:

peircean config wizard

Or by setting environment variables in a .env file:

# Provider Selection
PEIRCEAN_PROVIDER=anthropic  # anthropic, openai, gemini, ollama
PEIRCEAN_MODEL=claude-3-5-sonnet-20241022

# API Keys (Only needed for interactive/autonomous mode)
ANTHROPIC_API_KEY=sk-...
OPENAI_API_KEY=sk-...

# Feature Toggles
PEIRCEAN_ENABLE_COUNCIL=true
PEIRCEAN_INTERACTIVE_MODE=false # Set to true to let the tool call the LLM directly

📦 Supported Models

• Anthropic: Claude 3.5 Sonnet (Recommended), Claude 3 Opus, Claude 3 Haiku
• OpenAI: GPT-4o, GPT-4 Turbo
• Gemini: Gemini 1.5 Pro
• Ollama: Llama 3, Mistral

📚 Documentation

• Installation Guide
• Configuration Guide
• API Reference

📄 License

MIT