🎯 Goalie - AI Research Assistant

AI-powered research assistant with Goal-Oriented Action Planning (GOAP), advanced reasoning capabilities, and Model Context Protocol (MCP) integration for Claude Desktop.

Goalie combines intelligent search planning with the Perplexity API to deliver comprehensive research workflows. This Rust crate provides a native binary wrapper around the goalie npm package, offering seamless integration with both Rust and Node.js ecosystems.

🌟 What Makes Goalie Special?

• 🎯 GOAP Planning: Uses A* pathfinding algorithms for intelligent research workflows
• 🧠 Advanced Reasoning: Four specialized reasoning modes (Chain-of-Thought, Self-Consistency, Anti-Hallucination, Agentic Research)
• 🔧 MCP Integration: Native Claude Desktop integration with 11+ specialized tools
• 🔍 Smart Search: Enhanced Perplexity API integration with domain filtering and citation management
• 🛡️ Quality Assurance: Built-in fact-checking and hallucination prevention
• ⚡ Cross-Platform: Available as both Rust binary and npm package

Installation

Via Cargo (Rust)

cargo install goalie

Via npm (Node.js)

npm install -g goalie
# or run directly
npx goalie@latest --help

Quick Start

Set up your Perplexity API key:

# Get your API key from https://perplexity.ai/settings/api
export PERPLEXITY_API_KEY="pplx-your-key-here"
# Or add to .env file:
echo 'PERPLEXITY_API_KEY="pplx-your-key-here"' >> .env

Start researching immediately:

goalie search "What are the latest developments in quantum computing?"

Usage

All commands are forwarded to the npm package goalie@1.2.1:

Command Line Interface

# Interactive search with GOAP planning
goalie search "latest developments in renewable energy"

# Raw search without planning
goalie query "climate change impacts 2024"

# Advanced reasoning modes
goalie reason --mode chain-of-thought "Should companies invest in AI?"
goalie reason --mode self-consistency "Is nuclear energy safe?" --samples 5
goalie reason --mode anti-hallucination --claims "AI will replace all jobs" --citations "..."

# Multi-agent research
goalie research "future of space exploration" --agents researcher,analyst,critic

# MCP server for Claude integration
goalie start

# Configuration and diagnostics
goalie validate
goalie info
goalie --help

MCP Server Integration

Start the MCP server for Claude Desktop integration:

goalie start

Then add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "goalie": {
      "command": "goalie",
      "args": ["start"],
      "env": {
        "PERPLEXITY_API_KEY": "your-api-key-here"
      }
    }
  }
}

Available MCP Tools

• goap.search - Intelligent search with planning
• goap.plan.explain - Explain GOAP planning process
• search.raw - Direct Perplexity search
• reasoning.chain_of_thought - Multi-path reasoning analysis
• reasoning.self_consistency - Consensus-based verification
• reasoning.anti_hallucination - Fact-checking with citations
• reasoning.agentic_research - Multi-agent research workflows
• Plugin management tools (list, enable, disable, info)

Features

🎯 GOAP Planning System

Goal-Oriented Action Planning with A* pathfinding algorithm for intelligent research workflows:

• Automatic query decomposition and action planning
• Dynamic replanning based on search results
• Confidence-based path selection
• Multi-step reasoning chains

🔍 Enhanced Search Capabilities

Perplexity API integration with advanced features:

• Real-time web search with citations
• Domain filtering and recency controls
• Academic and web search modes
• Result synthesis and analysis

🧠 Advanced Reasoning Engine

Four specialized reasoning modes:

• Chain-of-Thought: Multi-branch exploration with confidence scoring
• Self-Consistency: Consensus validation across multiple samples
• Anti-Hallucination: Citation-based fact verification with token limits
• Agentic Research: Multi-agent collaborative analysis

🔧 Professional Integration

• MCP Protocol: Native Claude Desktop integration
• Plugin System: Extensible architecture with 11+ built-in tools
• CLI Interface: Complete command-line functionality
• API Access: Programmatic usage via npm package

🛡️ Quality Assurance

• Ed25519 cryptographic verification (experimental)
• Citation requirement enforcement
• Uncertainty flagging for unverified claims
• Token limit management to prevent overflow

Requirements

• Node.js 18+ (for the underlying npm package)
• Perplexity API Key (get one here)
• npm or npx available in PATH

Examples

Research Query

goalie search "state of the art in large language models 2024"

Multi-Agent Analysis

goalie research "impact of AI on healthcare" --agents researcher,analyst,critic --parallel

Fact Verification

goalie reason --mode anti-hallucination \
  --claims "ChatGPT was released in 2022" \
  --citations "OpenAI announced ChatGPT in November 2022"

Recent Updates

v1.2.1 (Latest)

• ✅ Fixed anti-hallucination tool token limit issues
• ✅ Added pagination support for large citation responses
• ✅ Improved MCP server stability
• ✅ All reasoning tools verified working with real API calls

v1.2.0

• ✅ Fixed mock reasoning handlers to use real Perplexity API
• ✅ Replaced server-side mock implementations
• ✅ Comprehensive reasoning tool verification

Publishing

The npm package is automatically published from the main repository:

# Update version
npm version patch|minor|major

# Publish to npm
npm publish

# Install latest version
npm install -g goalie@latest

License

MIT License - see LICENSE file for details.

Repository

https://github.com/ruvnet/goalie