Open OllaMan, and Ollama Starts Itself: Local Ollama in OllaMan 3.9

OllaMan is a client for Ollama, which means it has always had one awkward rule: Ollama has to be running first. Quit it by accident, reboot, or install it on a new machine, and the first thing you see in OllaMan is a connection error. You then leave the app, start Ollama, come back, and press refresh.

OllaMan 3.9 removes that detour. It can now find the Ollama on your computer, start it, stop it, install it if it is missing, and keep it up to date. You open OllaMan and get to work.

Local Ollama management is free for everyone. OllaMan does not bundle Ollama: when you ask it to install or update, it downloads the official release from Ollama's GitHub page.

Start and stop from the Dashboard

Look at the top right of the Dashboard. There is a small Ollama control with a status dot and a button.

When Ollama isn't running, the dot is red, the status says Not running, and the button says Start. One click and OllaMan brings Ollama up. When it is running, the same spot says Running and offers Stop. If Ollama isn't installed at all, the button says Install.

The rest of the Dashboard behaves sensibly while Ollama is down. The Running Models area says "Ollama is not running" instead of showing a red error, and Recently Installed says your models are unavailable, rather than telling you that you have none. Start Ollama and they all come back.

Stopping is guarded. OllaMan asks first, because it unloads every model and interrupts running chats and downloads.

OllaMan can also start Ollama for you. Start Ollama with OllaMan is on by default: when OllaMan opens with the local server selected and Ollama isn't running, it starts it. There is a matching Stop Ollama when OllaMan quits option, off by default, which only stops an Ollama that OllaMan itself started in that session.

It does it the way your install expects

Ollama isn't started the same way everywhere, and a client that simply launches its own copy would end up fighting whatever already looks after Ollama. So OllaMan starts and stops Ollama through the thing that owns your install:

• The Ollama app on macOS and Windows is started in the background, without a window. On macOS it is stopped with SIGTERM, because a normal quit only hides the app. On Windows its process tree is ended.
• Homebrew installs go through brew services.
• systemd on Linux goes through pkexec, so you get a normal password prompt. This is also why OllaMan never starts that service automatically.
• A bare ollama command with no app gets an ollama serve that OllaMan runs itself. It remembers it in a pid file, so it still knows it owns the process after you restart OllaMan.
• Anything else, such as ollama serve in a terminal or a Docker container, is shown as Another process, read-only. Stop it where you started it.

The program paths and arguments live in OllaMan's native code. The interface can pick an action, such as start or stop, but never a command line.

One-click install

If Ollama isn't there, click Install in the Dashboard header or Install Ollama in Settings.

OllaMan finds the latest release and downloads it. The download is often over 1 GB, so it can be cancelled and resumed, and it gives up with an error if the connection stalls for 60 seconds. Nothing is installed until the file passes two checks:

• Its SHA-256 checksum has to match.
• Its signature has to be Ollama's. On macOS that means codesign plus Ollama's Team ID. On Windows it means a valid Authenticode signature from Ollama Inc.

Then OllaMan installs Ollama and starts it. The status flips to Running and you can go straight to downloading a model.

On Linux, Ollama's official installer needs administrator rights, so OllaMan shows the install.sh command to copy into a terminal instead. After that, start and stop work from OllaMan as usual.

A settings page for your local Ollama

Everything above also lives in Settings → Local Ollama, together with the details.

The top of the page shows status, version, Managed by, install Location and Port, with Start or Stop right there. Below that is Server configuration.

It shows the model directory, the listen address and the context length, the three settings people most often wonder about, and tucks every other OLLAMA_* variable under All variables. OllaMan reads these from Ollama's own server log, or from the unit file on systemd. The section is read-only on purpose: it tells you what your Ollama is actually using, and a hint explains how to change it, either in the Ollama app's settings or with an environment variable such as OLLAMA_MODELS.

If OllaMan can't find Ollama, or picks the wrong copy, for example when Ollama lives on an external disk, use Ollama location → Choose… to point at the app or the ollama executable. OllaMan checks your choice and shows what it detected. Reset to automatic undoes it.

Update Ollama too

The same page checks the latest Ollama release (cached for six hours) and offers Update when yours is behind.

For the Ollama app on macOS and Windows, the update reuses the install checks, replaces the app in place, and restarts Ollama if it was running. If something goes wrong, the previous version is put back. Homebrew installs run brew upgrade ollama. Linux and other installs show the manual steps.

Updating restarts Ollama for a moment, which unloads loaded models and interrupts running chats and downloads, so OllaMan asks before it does it.

A Dashboard that says more

While we were in there, the Dashboard got a pass of its own.

• Metric cards share one layout: a title, a value with its unit, and a caption. The captions show which models are loaded, how they split between GPU and CPU, when your library last changed, and which model is the largest. The stray hover effect and the misleading "Online" badge are gone.
• Running models read like ollama ps: size, processor (100% GPU, or a split such as 20% CPU / 80% GPU), context length, and when the model unloads. The Unload button now has a label.
• Cards link where you expect: Installed and Disk go to the Installed page, and Recently Installed and Trending Now each have a View all link.

Try it

Existing installs get an auto-update prompt. To get 3.9 now, download it from ollaman.com, open the Dashboard, and look at the top right. If Ollama is already running there, you'll see it. If it isn't, you now know what to click. The Local Ollama guide covers every option in detail.

---