Ollama MCP Agent

Example Use Cases

• 1

  Run large language models locally without cloud costs and leverage Ollama or an alternative Gemini runtime.

• 2

  Call external tools or services through MCP servers (configured in mcp_config.json) to extend model capabilities like weather, fetch, or custom tools.

• 3

  Streaming responses and visibility into tool-call activity for debugging or monitoring during interactive sessions.

Description

This repository provides an MCP-compatible agent that runs LLM models locally via Ollama (with optional Google Gemini support) and lets you extend model capabilities by connecting MCP servers. It supports streaming output, tool-call information monitoring, and configurable MCP servers via mcp_config.json. It's useful for running LLMs offline on your PC with no per-request costs (Gemini requires a Google API key). The project includes example scripts, configuration, and instructions for installing dependencies and models.