Dive

Why we included this project

Dive is a desktop chat client that puts MCP-hosted tools and the models you already run into a single interface, which is handy if you juggle several LLMs or want to work with Model Context Protocol servers without hand-editing config files. It connects to ChatGPT, Anthropic, and Ollama, and because any OpenAI-compatible endpoint also works, a local model can slot in just as easily. A handful of built-in tools (web fetching, file management, shell commands) cover common agent tasks before you add your own servers, and the finer controls are refreshingly direct: individual MCP tools can be switched on or off, and API keys and models live in a simple settings file, so shortcuts and auto-saved drafts smooth the routine. It installs on Windows, macOS, and Linux with both a traditional Electron build and a lighter Tauri one, so it works as a dependable daily driver for individuals while giving developers a clear place to manage their own server connections.