Install and Serve

One install, one daemon, one port

Ollama installs a single binary and runs as a background daemon. After install, an HTTP server listens on localhost:11434. That's almost the whole picture. Whether the daemon starts by itself depends on how you installed it: the Linux script and the macOS app register and start a service for you, while a Homebrew install registers one and leaves starting it to you.

Install paths by platform

macOS — Homebrew is the cleanest path. brew install ollama installs the CLI and defines a service, but Homebrew never starts a service on your behalf. Follow it with brew services start ollama to run it now and at login, or ollama serve to run it in the foreground for one session. The macOS app from ollama.com is the other path, and that one does start on its own.
Linux — curl -fsSL https://ollama.com/install.sh | sh. Reads the script first if you don't trust pipe-to-shell; it installs systemd unit ollama.service.
Windows — official MSI installer from ollama.com.
Docker — for headless servers. Mount a volume so models persist across container restarts.

Verify the daemon is up

Once installed, the simplest health check is a GET on the root URL. If you see Ollama is running, the daemon is alive and you can move on. If you don't, the most common failure modes are: launchd / systemd failed to start, port 11434 is taken, or — the common one on macOS — Homebrew installed the formula and nobody ever ran brew services start ollama.

EV v3 · 2026-09-18: Two sentences in this lesson claimed a Homebrew install starts the daemon for you — “there's no separate server-start ritual” and “a launchd service that auto-starts the daemon”. That was wrong when it was written, not something Homebrew later changed: the formula page lists the service as brew services start ollama (“start now and at login”) or ollama serve to run it directly, and Homebrew has never auto-started a formula's service. The Linux half was right all along — install.sh really does systemctl enable and start the unit — which is probably how the macOS claim slipped through. If you installed by Homebrew and the health check at the bottom of this lesson never printed Ollama is running, that was why.

Code

Install + verify·bash

# macOS — Homebrew. The install does NOT start the daemon.
brew install ollama
brew services start ollama     # now and at login
# ...or, for one foreground session instead:
# ollama serve

# Linux — official install script
curl -fsSL https://ollama.com/install.sh | sh

# Docker — persistent volume
docker run -d \
  -v ollama:/root/.ollama \
  -p 11434:11434 \
  --name ollama ollama/ollama

# Verify the daemon (any platform)
curl -s http://localhost:11434
# expected: "Ollama is running"

# Check the version
ollama --version

Exercise

Install Ollama on your machine. On macOS with Homebrew, check curl http://localhost:11434 BEFORE starting the service and again after brew services start ollama — the difference is the whole lesson. Verify with curl http://localhost:11434 and ollama --version. If the daemon isn't running, identify why (launchd / systemd / port collision) and fix it before moving on.