Running Local LLMs with Ollama and Connecting Them to Coding Agents

Ollama lets users run open models such as Gemma, Qwen, DeepSeek, Kimi and gpt-oss on their own machines through a command-line tool, a local REST API, and official Python and JavaScript libraries. A guide walks through installing it on macOS, Linux, Windows, or Docker, then running a model in the terminal. The ollama launch command can connect coding agents like Claude Code, Codex, and OpenCode to local Ollama models. The guide notes that small local models are considerably weaker at agentic coding than frontier models, though they suit privacy-focused, offline, or low-cost experimentation. It also shows how to call a model from code via the REST API or client libraries, and notes that llama.cpp is the underlying backend for lower-level control.