LlamaBoss

Documentation

Getting started with LlamaBoss

LlamaBoss is a native Windows AI workspace. Run GGUF models locally, connect to an OpenAI-compatible endpoint when you choose, and give models controlled access to files and tools.

Applies to LlamaBoss v0.1.9 · Updated 2026-07-12

The LlamaBoss application showing a local AI chat with formatted code.
LlamaBoss combines local chat, files, projects, tools, and optional remote endpoints in one native Windows app.

Install

  1. Download the current Windows installer from the LlamaBoss home page.
  2. Run the .exe and complete the installer.
  3. Open LlamaBoss from the Start menu or desktop shortcut.
Windows SmartScreen

LlamaBoss is independently published and may not yet have enough reputation for SmartScreen. When the installer came from llamaboss.com, choose More info → Run anyway. The source is available on GitHub for inspection.

Choose your first model

Open the model pill at the top of the window or open Settings. You can download a curated model, select a GGUF already in your models folder, change the models folder, or select a model supplied by a configured remote endpoint.

A smaller model is the safest first test. It loads faster and makes it easier to confirm that the CPU or GPU runtime is working before you try a larger model.

Start a chat

  1. Type a message and press Enter.
  2. Use Shift + Enter for a new line.
  3. Use the paperclip, drag-and-drop, or clipboard paste to add supported files.
  4. Turn on the robot button when the task needs Agent Mode tools.
A useful first test

Ask the model to explain a small text file or screenshot. Then enable Agent Mode and ask it to list the current workspace. This tests ordinary chat, attachments, and a safe read-only tool.

Explore the features

Interface & conversationsSidebar search, project groups, multiple windows, and chat shortcuts. Models & performanceGGUF models, vision companions, context length, KV cache, and VRAM. Files & documentsImages, code, PDFs, Word, Excel, CSV, and ZIP workflows. Agent Mode & approvalsTool categories, workspaces, approval cards, and stopping work. Projects, Workflows & SkillsShared context, source files, repeatable plans, and reusable capabilities. GoalsGive a larger objective, let LlamaBoss continue, and verify completion. Remote endpointsOpenAI-compatible services, Connections, tool protocols, and image models. Commands & shortcutsA searchable reference for power-user controls.

Local-first does not mean local-only

With a local GGUF model, conversation inference stays on your computer and can work offline after the model is downloaded. LlamaBoss can also use network features that you deliberately invoke: model downloads, update checks, webpage retrieval, package installation, and configured remote endpoints.

See Privacy & data for exact storage locations and network behavior.

Interface & conversations →