Moose 0.8.0
Web search and calculations join your chats, OpenRouter joins the provider list, and finding a model gets easier.
Chat with local and cloud AI on Linux.
Ask questions, work on your writing and talk through code in a native Linux app. Run models on your computer with Ollama, connect your own server, or choose an online service such as OpenRouter. Bring your files into the conversation, look things up with a compatible model, and keep your chats and drafts on your device.
flatpak install flathub io.github.moooossee.MooseRequires Flatpak and the Flathub remote. Install Moose, then set up Ollama or connect a provider from the app. Getting started

Look things up, check calculations, connect to OpenRouter and find a model for what you want to do. Longer answers can now continue where they stopped.
Images require a model with vision support. PDFs need selectable text. See Files and documents.
Open Models and choose Everyday, Coding, Images or Documents. Narrow the list by download size, image input, tool support or multiple languages, and compare versions before choosing one.
Model details show estimated memory and storage needs. When Moose manages Ollama, suggestions also consider the memory available on your computer. These are estimates to help you choose; longer chats and images can need more memory. Choose Use beside an installed model to start a chat with it.
Look something up or check a calculation without leaving your conversation. With a compatible Ollama or Ollama Cloud model, turn on Web or Calculator in Chat Settings, or type /web or /calculator. You can see what the tools did and read their sources alongside the answer.
Web starts with DuckDuckGo, without a search account. Choose Brave Search, SearXNG or Links Only in Preferences → Web. Web asks for separate permission before first use and is blocked by Local Only mode. Calculator runs on your device; its results are shared with your chat's model.
Choose Add Model… in Models and paste a Hugging Face repository link. Leave the variant on Automatic, or choose a quantization such as Q4_K_M or an exact GGUF filename. Moose shows the model source and the Ollama connection that will download it.
You can also paste a model link into model search. Installed Hugging Face models link back to their repository, and an already installed chat model can open a new conversation straight from the download dialog.
Choose a GGUF model supported by your Ollama version and suited to the memory available on the device running it.
Use Ollama managed by Moose, connect an existing Ollama instance, or bring your own credentials for a cloud provider:
If your service is not listed, add a custom connection using a compatible API. Moose can find available models or let you add them by name. Images and reasoning options depend on the model you choose.
Local Only mode uses Moose-managed Ollama and blocks cloud connections, external Ollama and web tools. Remote chats ask permission before sharing messages, and files need separate permission. Keys are kept in your desktop's secure key storage.
Open Preferences → Ollama & Diagnostics to check for engine updates or repair an installation. Interrupted engine downloads can resume, and a failed update restores the previous installation. Your models, chats and connections stay put.
If a model needs a newer Ollama version, Moose can update its managed installation and retry the download. You can also view recent logs and copy error details when you need help.
Web search and calculations join your chats, OpenRouter joins the provider list, and finding a model gets easier.