← Back to Discover

Ollama

Open-source tool for running open models such as Gemma, Qwen and DeepSeek from the command line, with a local REST API, Python and JavaScript libraries, and installers for macOS, Windows, Linux and Docker.

Local modelsOpen modelsCommand-line tools
Visit website ↗

What it does

Ollama is an open-source project for running large language models. The repository says it gives access to open models locally or in the cloud with privacy as a stated aim. You install it with a one-line script on macOS and Linux, a PowerShell command on Windows, or the official Docker image, and then start a model with a command such as ollama run gemma4. Applications talk to it through a REST API on localhost, shown at port 11434, and official Python and JavaScript libraries wrap that API. The ollama launch command connects Ollama to coding tools such as Claude Code, Codex, Copilot CLI and OpenCode, and can set up the OpenClaw personal assistant across WhatsApp, Telegram, Slack and Discord. The README also carries a long list of community integrations, including chat interfaces, desktop and mobile clients, editor plugins, SDKs for many languages, RAG tools, bots and monitoring tools. It lists llama.cpp as a supported backend, and the project is released under the MIT license.

Capabilities

  • Run open models from the command line, for example ollama run gemma4
  • REST API on localhost for running and managing models
  • Official Python and JavaScript libraries
  • ollama launch to connect tools such as Claude Code, Codex and OpenCode
  • Install on macOS, Windows, Linux or with the official Docker image
  • Large list of community integrations, from chat UIs to SDKs and RAG tools

When to use it

  • You want to run open models on your own machine rather than use a hosted chatbot
  • You want a local model endpoint that scripts and apps can call over REST or a client library
  • You want to point a coding agent at models served through Ollama