Modelo
Modelo is a native macOS LLM client that i work on with a friend.
i run models on different machines and different cloud providers as well. i want those backends in one application, without treating each server as a separate chat product.
One client, several backends
Modelo connects to local runtimes such as llama.cpp, Ollama, and oMLX, as well as openai-compatible cloud APIs. it combines streaming chat with model selection, context tracking, usage reports, and server status.
larger model outputs can open in an artifact panel with previews for HTML, SVG, Mermaid, and Markdown. The project also has opt-in tools and MCP support.
My work on it
i’ve been focused on the things i encounter while using it:
- keeping replies running when i switch conversations, with completion notifications.
- making tool-round limits configurable.
- adding context-length overrides for backends that do not report them.
- improving model and server reporting.
- working through local builds, signing, and installation.
- reviewing changes around tools, memory, and cloud access.
Modelo is shared work, not a solo project. it is also a useful place to turn the rough edges of my local-model setup into improvements other people can use.