Claude Desktop API Integration via MCP

What is Claude Desktop API Integration via MCP?

This project develops an MCP (Machine Communication Protocol) server that integrates Claude Desktop with the Claude API. It allows users to access advanced features such as system prompts, conversation management, and longer context windows directly from Claude Desktop.

How to use Claude Desktop API Integration via MCP?

Clone the repository, install dependencies with pip install -r requirements.txt, copy .env.example to .env and set your ANTHROPIC_API_KEY. Then configure Claude Desktop by editing claude_desktop_config.json (located at ~/Library/Application Support/Claude/ on macOS or %APPDATA%\Claude\ on Windows) with the server path. Use @claude-api in chat to invoke API tools (e.g., @claude-api query_claude prompt="What is the capital of France?").

Key features of Claude Desktop API Integration via MCP

• Direct connection to Claude API through MCP
• Conversation history tracking and management
• System prompt support
• Seamless switching between Professional Plan and API usage
• Support for latest Claude models (Opus, Sonnet, Haiku)
• FastAPI service for external access

Use cases of Claude Desktop API Integration via MCP

• Use the API for longer context windows beyond the Professional Plan limit
• Employ custom system prompts to guide Claude’s behavior (e.g., fitness coach)
• Bypass rate limits or manage multiple conversation threads with IDs
• Switch between free Professional Plan chats and API‑powered queries on demand

FAQ from Claude Desktop API Integration via MCP

How do I switch between Professional Plan and API usage?

Use normal chat for Professional Plan. Prefix any message with @claude-api to route the request through the MCP server and use your API key and credits.

What models are supported?

The server supports Claude 3 Opus, Claude 3 Sonnet, Claude 3 Haiku, and Claude 3.5 Sonnet. Use @claude-api list_available_models to see all available models.

How do I manage conversations?

Use @claude-api query_claude with a conversation_id parameter to track a thread. Retrieve history with get_conversation_history, list all conversations with list_conversations, and clear one with clear_conversation.

Does using the API cost money?

Yes, API calls use your Anthropic API credits and may incur charges. It is recommended to use the Professional Plan for general inquiries and @claude-api only when you need advanced features.

Where is my data stored?

Conversation histories are managed server‑side by the MCP server; no data is stored outside your local environment unless you configure otherwise. The server itself runs locally.