claude-command 0.5.0

Claude Command - AI command center for Claude Code

Claude Command MCP Server

Real-time AI command center for Claude Code CLI with dynamic multi-provider support.

[!IMPORTANT] System Prompt: Development standards and procedures are documented in .air/prompts/system.prompt.md - this is the most important file for understanding how to work with this codebase.

Overview

Claude Command is an MCP (Model Context Protocol) server that enables live three-way conversations between Claude Code CLI, multiple AI providers, and human developers. Built with fastMCP, it supports dynamic provider switching and real-time streaming conversations.

Features

• Multi-Provider Support: Gemini, OpenAI, and Anthropic models
• Dynamic Provider Switching: Call any AI provider on-demand
• Real-Time Streaming: Watch conversations unfold live via tail -f
• Context Preservation: Full conversation history maintained across providers
• Background Processing: Non-blocking AI calls with threading
• Session Management: Persistent conversation storage in JSON format

Architecture

Built using fastMCP framework with modular client architecture and mission-based execution:

Claude Code CLI (Hub) ↔ fastMCP Server ↔ Mission System ↔ [Gemini ↔ OpenAI ↔ Anthropic]
                                        ↓
                        [Query | Review | Brainstorm | Conversation | Recon]
                                        ↓
                           [Streaming | Sessions | File Management]

Mission-Based Architecture

The core functionality is organized into specialized missions that handle different types of AI interactions:

• Query Mission: Multi-provider intelligence gathering with independent responses
• Review Mission: Code inspection across multiple AI providers with focus areas
• Brainstorm Mission: Creative collaboration with multiple providers simultaneously
• Conversation Mission: Live conversations with context preservation and history
• Recon Mission: Independent reconnaissance for unbiased comparative analysis

Installation

Prerequisites

• Claude Code CLI
• Python 3.10+
• API keys for desired providers

Install Dependencies

git clone https://github.com/shaneholloman/mcp-claude-command
cd mcp-claude-command
uv sync

Configure API Keys

Set environment variables:

export GEMINI_API_KEY="your_gemini_key"
export OPENAI_API_KEY="your_openai_key"
export ANTHROPIC_API_KEY="your_anthropic_key"

Or use command line arguments:

uv run python -m claude_command \
  --gemini-api-key YOUR_GEMINI_KEY \
  --openai-api-key YOUR_OPENAI_KEY \
  --anthropic-api-key YOUR_ANTHROPIC_KEY

Add to Claude Code

Install as a global UV tool and configure with Claude CLI:

# Install claude-command globally
uv tool install .

# Configure with Claude CLI (includes all API keys for multi-provider support)
# Option 1: Using environment variables (if you have them set)
claude mcp add claude-command claude-command \
  -e GEMINI_API_KEY="$GEMINI_API_KEY" \
  -e CLAUDE_COMMAND_OPENAI_API_KEY="$OPENAI_API_KEY" \
  -e CLAUDE_COMMAND_ANTHROPIC_API_KEY="$ANTHROPIC_API_KEY" \
  --scope user

# Option 2: Using your actual API keys directly (replace with your real keys)
claude mcp add claude-command claude-command \
  -e GEMINI_API_KEY="AIzaSyBxxxxxxxxxxxxxxxxxxxxxxx" \
  -e CLAUDE_COMMAND_OPENAI_API_KEY="sk-proj-xxxxxxxxxxxxxxxxxxxxxxx" \
  -e CLAUDE_COMMAND_ANTHROPIC_API_KEY="sk-ant-api03-xxxxxxxxxxxxxxx" \
  --scope user

# Verify configuration
claude mcp list
claude mcp get claude-command

# Verify API keys are properly configured (incontrovertible proof)
claude mcp get claude-command | grep -A 10 "Environment:"

[!IMPORTANT] Now, restart Claude's CLI. This is important.

Usage

Available Tools

Tool	Description	Parameters
claude_command_query	Interactive conversation or multi-provider query	prompt, conversation_id, temperature, provider, providers
claude_command_review	Code review by single or multiple AI providers	code, focus, provider, providers, temperature
claude_command_brainstorm	Creative brainstorming with single or multiple providers	topic, context, provider, providers, temperature
claude_command_recon	Independent reconnaissance across multiple providers	survey_prompt, providers, temperature
claude_command_conversation_history	Retrieve conversation history	conversation_id
claude_command_status	Server health and configuration	-
claude_command_conversations_list	List all conversation files	-
claude_command_conversations_cleanup	Clean up old conversations	keep_count

Provider Selection

Specify AI provider for any tool:

# Use default provider (first available)
claude_command_query prompt="Hello!"

# Use specific provider
claude_command_query prompt="Hello!" provider="gemini"
claude_command_query prompt="What do you think?" provider="openai"
claude_command_query prompt="Your perspective?" provider="anthropic"

Live Streaming

Watch conversations in real-time:

tail -f ~/claude-command/live_stream.txt

Multi-AI Conversations

Create conversations between multiple AI providers:

1. Start conversation with one provider
2. Use different providers for follow-up responses
3. Each AI sees full conversation context
4. Watch live streaming for real-time collaboration

Mission System

The Claude Command server is built around a mission-based architecture where each mission type handles specific AI interaction patterns. This system provides specialized functionality while maintaining consistent streaming, session management, and multi-provider support.

Mission Types

1. Query Mission (claude_command_query with providers parameter)

Purpose: Multi-provider intelligence gathering with independent, unbiased responses.

Key Features:

• Each provider receives identical prompts with zero context
• Parallel execution with real-time streaming
• Independent responses prevent cross-contamination
• Ideal for comparative analysis and diverse perspectives

Usage:

# Query all providers
claude_command_query prompt="What are the pros and cons of microservices?" providers="gemini,openai,anthropic"

# Query specific providers
claude_command_query prompt="Best practices for API design" providers="gemini,openai"

Output Structure:

• Individual provider stream files: query-TIMESTAMP-PROVIDER.md
• Consolidated summary file: query-TIMESTAMP-summary.md
• JSON results: ~/claude-command/queries/query-TIMESTAMP.json

2. Review Mission (claude_command_review)

Purpose: Code inspection across multiple AI providers with structured focus areas.

Key Features:

• Specialized code analysis prompts
• Multiple focus areas: general, security, performance, best practices
• Multi-provider or single-provider modes
• Formatted code review output

Usage:

# Multi-provider code review
claude_command_review code="def process_data(data): return [x*2 for x in data]" focus="performance" providers="gemini,openai,anthropic"

# Single provider review
claude_command_review code="..." focus="security" provider="anthropic"

Focus Areas:

• general: Overall code quality and structure
• security: Security vulnerabilities and best practices
• performance: Optimization opportunities and bottlenecks
• readability: Code clarity and maintainability
• testing: Test coverage and testability

3. Brainstorm Mission (claude_command_brainstorm)

Purpose: Creative collaboration with multiple AI providers for diverse problem-solving perspectives.

Key Features:

• Higher temperature settings for creativity
• Context-aware brainstorming
• Independent ideation across providers
• Collaborative problem-solving approach

Usage:

# Multi-provider brainstorming
claude_command_brainstorm topic="Improving user onboarding" context="SaaS application with 10k users" providers="gemini,openai,anthropic"

# Single provider brainstorming
claude_command_brainstorm topic="Cost optimization strategies" provider="gemini"

4. Conversation Mission (claude_command_query without providers parameter)

Purpose: Live conversations with context preservation and full conversation history.

Key Features:

• Maintains conversation context across exchanges
• Session-based conversation management
• Single provider focus for consistency
• Real-time streaming with history preservation

Usage:

# Continue existing conversation
claude_command_query prompt="Can you elaborate on that?" conversation_id="session_2025-06-15_12-30-45_abc123def"

# Start new conversation
claude_command_query prompt="Help me understand microservices architecture" provider="anthropic"

5. Recon Mission (claude_command_recon)

Purpose: Independent reconnaissance for unbiased comparative intelligence gathering.

Key Features:

• Each provider maintains separate conversation history
• Prevents cross-contamination between providers
• Specialized for reconnaissance and surveying
• Individual provider conversation context

Usage:

# Survey all providers independently
claude_command_recon survey_prompt="What are emerging trends in AI development?" providers="gemini,openai,anthropic"

# Targeted recon with specific providers
claude_command_recon survey_prompt="Current state of quantum computing" providers="openai,anthropic"

Mission Execution Flow

1. Mission Selection: Based on tool called and parameters provided
2. Provider Determination: Single provider vs. multi-provider execution
3. Parallel Processing: Multi-threaded execution for multi-provider missions
4. Real-time Streaming: Live output to provider-specific stream files
5. Result Compilation: Aggregated results with metadata and file references
6. Session Management: Conversation history and file organization

Streaming Infrastructure

Each mission utilizes the comprehensive streaming system:

File Structure:

~/claude-command/
├── live-streams/           # Real-time streaming files
│   ├── query-TIMESTAMP-gemini.md
│   ├── query-TIMESTAMP-openai.md
│   ├── query-TIMESTAMP-anthropic.md
│   └── query-TIMESTAMP-summary.md
├── conversations/          # Conversation sessions
├── queries/               # Query mission results
├── reviews/               # Code review results
├── brainstorm/           # Brainstorming sessions
└── recon/                # Reconnaissance results
    └── conversations/    # Provider-specific recon history

Live Streaming Features:

• Word-by-word streaming simulation
• Provider-specific stream files with metadata
• YAML front matter with operation details
• Tmux multi-pane viewing commands
• Thread-safe file operations

Provider Management

Dynamic Provider Selection:

• provider="gemini" - Single provider execution
• providers="gemini,openai" - Multi-provider execution
• Default: All available providers

Provider Fallback:

• Automatic fallback to available providers
• Error handling for unavailable providers
• Provider-specific error reporting

File Organization

Directory Structure:

• conversations/: Live conversation sessions
• queries/: Multi-provider query results
• reviews/: Code review analyses
• brainstorm/: Creative brainstorming sessions
• recon/: Independent reconnaissance data
• live-streams/: Real-time streaming files

File Naming Convention:

• MISSION-TIMESTAMP-PROVIDER.md - Individual provider streams
• MISSION-TIMESTAMP-summary.md - Consolidated summaries
• MISSION-TIMESTAMP.json - Structured result data

File Locations

All data organized in mission-specific directories under ~/claude-command/:

Mission Data Structure

• Live Streams: live-streams/ - Real-time streaming files for all missions
• Conversations: conversations/ - Live conversation sessions with history
• Queries: queries/ - Multi-provider query results and analysis
• Reviews: reviews/ - Code review outputs and analysis
• Brainstorm: brainstorm/ - Creative brainstorming session results
• Recon: recon/ - Independent reconnaissance data and provider histories

File Types

• Stream Files: MISSION-TIMESTAMP-PROVIDER.md - Real-time provider outputs
• Summary Files: MISSION-TIMESTAMP-summary.md - Consolidated mission summaries
• Result Files: MISSION-TIMESTAMP.json - Structured mission results
• Session Files: session_YYYY-MM-DD_HH-MM-SS_XXXXXXXX.json - Conversation sessions
• Recon History: recon/conversations/recon-PROVIDER-conversation.json - Provider-specific histories

Configuration

Settings

Configure via environment variables or command line:

• GEMINI_API_KEY / --gemini-api-key
• OPENAI_API_KEY / --openai-api-key
• ANTHROPIC_API_KEY / --anthropic-api-key
• CONVERSATIONS_DIR / --conversations-dir (default: ~/claude-command)
• DEFAULT_TEMPERATURE / --temperature (default: 0.7)

Models

Default models by provider:

• Gemini: gemini-2.0-flash-exp
• OpenAI: gpt-4
• Anthropic: claude-3-5-sonnet-20241022

Development

Project Structure

src/claude_command/
├── server.py              # Main MCP server with fastMCP tools
├── settings.py            # Configuration management
├── sessions.py            # Session and file management
├── streaming.py           # Multi-file streaming infrastructure
├── missions/              # Mission-based execution system
│   ├── __init__.py       # Mission module exports
│   ├── query.py          # Multi-provider intelligence gathering
│   ├── review.py         # Code inspection across providers
│   ├── brainstorm.py     # Creative collaboration system
│   ├── conversation.py   # Live conversations with history
│   └── recon.py          # Independent reconnaissance missions
└── clients/              # AI provider implementations
    ├── interface.py       # AI client interface
    ├── gemini.py         # Gemini implementation
    ├── openai.py         # OpenAI implementation
    ├── anthropic.py      # Anthropic implementation
    └── factory.py        # Client factory and management

Local Development

git clone https://github.com/shaneholloman/mcp-claude-command
cd mcp-claude-command
uv sync
uv run python -m claude_command --gemini-api-key YOUR_KEY

Requirements

• Claude Code CLI (will NOT work with Claude Desktop)
• At least one API key (Gemini, OpenAI, or Anthropic)
• Python 3.10+
• uv package manager

License

MIT License - see LICENSE for details.