Repository navigation
LlamaRun for Windows, early access: attachments, memory, skills and much more.
Install: download LlamaRun-Setup-0.5.0-x64.exe below and run it. The installer isn't code-signed yet, so Windows SmartScreen may say "Windows protected your PC"; choose More info → Run anyway.
What's new
- Attach images and documents: use the paperclip, paste, or drag files into the chat window or the overlay. PDFs, Word documents and text files are read; images go to vision models.
- Diagrams and live previews in chat: flowcharts and other Mermaid diagrams draw right in the reply, SVGs show as pictures, and HTML becomes a live page that grows to fit its content.
- Memory: the AI remembers what you tell it about yourself between chats. See, edit or clear it in Settings → Memory.
- Skills: add SKILL.md skills (the same format Claude uses) from a ZIP, a folder or a link, or let the AI install one by following its README.
- Skills live in the shared
~/.agents/skillsfolder other AI apps read too, and your Claude Code skills show up automatically. - A skill's scripts ask before they run.
- Skills live in the shared
/menu: type/to run any skill, plugin command,/compactor/new.- Long chats: older messages are summarised automatically instead of being dropped, or type
/compact. Each model uses its own context size. - More powerful plugins: plugins can add
/commands, post rich messages, add context to every chat, use memory and skills, and re-theme the app. - Fixes: the overlay stays open while you pick a file, and right-clicking text gives spelling suggestions, Cut, Copy and Paste.
Still in 0.5.0
- Press Ctrl + Shift + A in any app to ask a question; Esc hides it. You can change the hotkey in Settings.
- Local models through Ollama, or cloud models after signing in with Google. Pro is free during early access.
- MCP servers and Python plugins, including Plugin Creator and the EDITH voice assistant.
64-bit Windows only.
SHA-256 LlamaRun-Setup-0.5.0-x64.exe: 80b7f043ddedf6e9e3f6d80e497dd89135b258e4725bb5cb7e224ce4000b5269