This CLI tool enables users to interact with Large Language Models (LLMs) and Model Context Protocol (MCP) compatible servers directly from their terminal. It supports various LLM providers (OpenAI, Groq, local models) and integrates with MCP servers for functionalities like web search and YouTube summarization, offering a versatile alternative to desktop clients.

How It Works

The client leverages a configuration file (~/.llm/config.json) to manage LLM provider details, system prompts, and MCP server integrations. It supports piping text and image input, using prompt templates for common tasks, and dynamically calling external tools (MCP servers) based on user queries. Tool execution can be confirmed manually or bypassed with a flag, and intermediate tool messages can be suppressed for cleaner script output.

Quick Start & Requirements

• Install via pip: pip install mcp-client-cli
• Requires a ~/.llm/config.json file for LLM and MCP server configuration.
• Optional: pngpaste for image clipboard support on macOS (brew install pngpaste).
• Optional: xclip for clipboard support on Linux (sudo apt install xclip).
• Official documentation: CONFIG.md

Highlighted Details

• Supports piping images for multimodal LLM analysis.
• Integrates with MCP servers via uvx or npx for tools like Brave Search and YouTube summarization.
• Offers prompt templates for common tasks (e.g., git review, commit message generation).
• Includes clipboard support for text and images across Windows, macOS, and Linux (including WSL).

Maintenance & Community

• Open to issues and pull requests for improvements.

Licensing & Compatibility

• License: MIT
• Compatible with commercial and closed-source applications.

Limitations & Caveats

Clipboard support on Linux requires the xclip utility to be installed. Image handling via clipboard on macOS is optional and requires pngpaste.