Ollama runs open-weight large language models on your own machine and serves them over a local REST API. It runs on macOS, Windows, and Linux, and an official Docker image is published alongside the native installs.
Starting ollama prompts you either to run a model or to connect Ollama to agents and applications you already use. ollama launch points a supported coding agent at local models without changing the way you work with it, with documented integrations for Claude Code, Codex, Copilot CLI, DeepSeek Harness, Droid, and OpenCode. ollama launch openclaw runs OpenClaw on Ollama as a personal assistant reachable from WhatsApp, Telegram, Slack, Discord, and other channels.
Models are pulled and run by name from the Ollama library, for example ollama run gemma4. A Modelfile describes how a model is packaged and parameterized, and existing weights can be imported. Inference is backed by the llama.cpp project.
A local REST API covers chat and model management, with official Python and JavaScript client libraries wrapping it. Alongside local execution, the project offers hosted cloud models on paid usage tiers, hosted in the United States, Europe, and Singapore, with local models remaining free and never leaving the machine.
A large community integration list covers chat interfaces, editors, libraries, agent frameworks, retrieval-augmented generation stacks, messaging bots, and terminal clients, including Open WebUI, Cline, Continue, LangChain, LlamaIndex, Haystack, LiteLLM, crewAI, and RAGFlow.
Features
- Local model serving: open-weight models run on your own hardware behind a local REST API
- Cross-platform: native installs for macOS, Windows, and Linux, plus an official Docker image
- Coding agent integrations: ollama launch starts Claude Code, Codex, Copilot CLI, DeepSeek Harness, Droid, or OpenCode against local models
- Assistant mode: OpenClaw runs on Ollama across WhatsApp, Telegram, Slack, Discord, and more
- Model library: models are pulled and run by name from a published library of open models
- Modelfile: model packaging and parameters are declared in a Modelfile, and external weights can be imported
- Client libraries: official Python and JavaScript packages, plus community SDKs for Java, .NET, Go, Rust, Swift, Ruby, PHP, and more
- llama.cpp backend: inference runs on the llama.cpp project
- Cloud tier: paid hosted capacity serves the same open models, with local use free and local data staying on the machine
- Ecosystem integrations: community hookups for Open WebUI, LangChain, LlamaIndex, Haystack, Cline, Continue, crewAI, RAGFlow, and many others
