MCP Recon Client

What is MCP Recon Client?

MCP Recon Client is a Model Context Protocol (MCP) client that uses open‑source LLM models to communicate with MCP servers, giving the LLM access to tools exposed by those servers. It is designed for developers and users who want to experiment with local or cloud‑based LLMs in the MCP ecosystem.

How to use MCP Recon Client?

Clone the repository, install dependencies with uv pip install -r requirements.txt, and install Ollama (ollama.com) if you plan to use local open‑source models. Optionally obtain a free Google Gemini API key and save it in the .env file. Then run the client with uv run src/main.py. The default main.py implementation uses Google Gen AI; alternative client implementations are available for other backends.

Key features of MCP Recon Client

• Integrates with Ollama to run open‑source LLMs locally.
• Supports Google Gemini via a free API key (optional).
• Provides tool access from MCP servers to the LLM.
• Includes multiple client implementations for different models.
• Managed with UV for Python dependency handling.

Use cases of MCP Recon Client

• Enabling open‑source LLMs to call tools exposed by MCP servers.

FAQ from MCP Recon Client

What LLM models does MCP Recon Client support?

It supports any open‑source model pulled through Ollama, as well as Google Gemini models when a Google API key is provided. The default main.py uses google‑gemini‑2.5‑pro.

What are the runtime requirements?

Python, the UV package manager, and either Ollama (for local models) or a Google Gemini API key. The repo also requires cloning and dependency installation.

Is MCP Recon Client fully tested for tool calling?

No. The author notes that comprehensive testing for suitable tool‑calling LLMs was not performed, and results heavily depend on correct prompting. The Google Gemini implementation gave the best observed results due to its large context window.

Can I use other LLM models besides Google Gemini?

Yes. The repository contains alternative chat client implementations that can be used with other models after adjusting the source code and pulling those models via Ollama.