Demonstrates integrating AI agents with local and remote tool execution via MCP and Ollama.
This project provides a comprehensive demonstration of leveraging AI agents, specifically using `smolagents`, to interact with a Multi-tool Control Protocol (MCP) server. It showcases both local tool execution within a GitHub Codespaces environment and integration with GitHub's new remote MCP server for cloud-based tool management. Furthermore, the toolkit illustrates how to host multiple tools from a single endpoint using `FastMCP` and how large language models (LLMs) powered by Ollama can intelligently select and utilize appropriate tools based on prompts, enabling dynamic and extensible agent capabilities.
Key Features
01Enables multi-tool hosting from a single endpoint using FastMCP
02Supports GitHub’s Remote MCP Server for cloud-based tool management
030 GitHub stars
04Demonstrates intelligent tool selection by LLMs via Ollama
05Integrates with local MCP Inspector in GitHub Codespaces
06Provides examples for both CLI and chat-based agent interactions
Use Cases
01Developing and testing AI agents that leverage external tools
02Experimenting with local and cloud-based tool orchestration for agentic workflows
03Creating robust agentic applications requiring dynamic access to specialized functionalities