ollama

Ollama is a tool for running open large language models on your own machine. It's built for developers who want to run models locally rather than calling a hosted API, whether for privacy, cost, offline use, or experimentation. Beyond running models directly, Ollama also connects to existing coding agents and assistants like Claude Code, Codex, Copilot CLI, Droid, and OpenCode, and it can power a personal AI assistant across WhatsApp, Telegram, Slack, and Discord through the OpenClaw integration.

Key features

• One-line install scripts: install with a single shell command on macOS, Windows, or Linux, or via the official Docker image on Docker Hub.
• Simple model management: pull and run models from the Ollama library with a single command, such as ollama run gemma4.
• REST API: a local HTTP API for chat and generation that other applications can call directly, running on localhost:11434 by default.
• Official client libraries: Python and JavaScript libraries (ollama on PyPI and npm) for calling models from application code.
• Coding agent integrations: launch supported coding tools directly, for example ollama launch claude, with documented integrations for several CLI coding agents.
• llama.cpp backend: built on the llama.cpp inference engine, giving it a widely used and actively maintained foundation for running models efficiently.
• Wide ecosystem support: listed as the default or first-class local backend by a large number of community chat interfaces, code editors, RAG tools, agent frameworks, and observability platforms.

Ideal use cases

Ollama fits developers who want to prototype or run production workloads against local models without depending on a hosted API, and teams that need models running entirely on their own infrastructure for data privacy or offline requirements. It's also useful as a drop-in backend for the large ecosystem of community chat interfaces, RAG tools, and agent frameworks that already support it, since many of these projects list Ollama as their default or first-class local backend, from web UIs like Open WebUI to terminal tools and IDE extensions. Developers wiring local models into coding agents or personal assistants benefit from the built-in integrations rather than configuring API endpoints by hand.

It's a weaker fit if you need the largest frontier models with no local hardware to run them, since local inference is bounded by the memory and compute you have available. If your workload requires a fully managed, auto-scaling hosted endpoint, a cloud model provider will be simpler than self-managing local infrastructure, and very large models may simply not fit on typical consumer hardware.

The community integration list is organized into categories including chat interfaces for web, desktop, and mobile, code editor extensions, agent and orchestration frameworks, RAG and knowledge base tools, terminal and CLI clients, observability platforms, and infrastructure options for deploying on cloud providers or through package managers like Homebrew and Pacman. That breadth means most existing AI tooling in a given category likely already has some path to running against Ollama rather than requiring custom integration work.

Installation

On macOS or Linux:

curl -fsSL https://ollama.com/install.sh | sh

On Windows (PowerShell):

irm https://ollama.com/install.ps1 | iex

Or use the official Docker image, ollama/ollama, available on Docker Hub. After installing, run ollama to get started, or run a specific model directly:

ollama run gemma4

To use it from Python:

pip install ollama

from ollama import chat
response = chat(model='gemma4', messages=[{'role': 'user', 'content': 'Why is the sky blue?'}])
print(response.message.content)

A similar ollama package is available for JavaScript via npm, and full CLI and REST API references, along with a guide for importing custom models and building from source, are documented on the project's docs site.