Run powerful AI models on your own machine—or connect to an OpenAI-compatible endpoint when you choose. Native C++, multimodal, and agentic.
Bundled with llama.cpp runtimes — the same engine that powers local AI everywhere. No Docker, no Python, no external dependencies.
Drop in or paste images, code, PDFs, Word documents, spreadsheets, CSV files, and ZIP archives.
Read-only inspections can run automatically; commands and mutating actions pause for clear approval.
Local GGUF inference stays on your PC. Remote endpoints and network tools are used only when you choose them.
LlamaBoss can take on multi-step jobs, reuse project context, work across documents, and verify larger objectives—without turning the desktop app into an opaque autonomous process.
Open several chats while one application-owned local-model service prevents duplicate VRAM loads.
Keep durable context and sources with a Project, repeat project-specific plans with Workflows, and reuse global Skills anywhere.
Inspect or create PDFs, Word documents, Excel workbooks, CSV reports, source archives, and reviewable scripts.
Define success criteria, continue across tool turns, collect evidence, and run a separate completion check.
Local models and saved work stay on your PC. Optional remote endpoints, model downloads, update checks, webpage retrieval, and package installation are clearly separated and used only when the selected feature needs them.
Bundled llama.cpp CPU and NVIDIA CUDA runtimes. No Docker, Python environment, Ollama service, or account is required for local chat.
Low-risk inspection can stay fluid. File mutations, scripts, installs, archive extraction, and broader commands pause for review.
Use curated downloads, your own GGUF folder, or configured OpenAI-compatible services from the same model picker.
Projects, Skills, notes, workflows, and generated artifacts live in visible Windows folders that you can inspect and back up.
Windows installer, MIT licensed, no account needed. Open the installer, click through, and start chatting with your local models.
Beta This native Windows build ships with bundled llama.cpp runtimes — no Ollama required. Download, install, and start chatting with local models right away.
