witsy

Why we included this project

Witsy sits on your desktop and talks to whatever model you already have keys for, or to none at all if you point it at a local Ollama instance. It stays deliberately neutral across providers, so you can mix hosted chat from OpenAI, Anthropic, or DeepSeek with free local models in the same app. What sets it apart is the Model Context Protocol support: you can attach MCP servers even when the underlying model runs locally, which most MCP clients won't let you do. On top of chat it handles image and video generation, text-to-speech, speech-to-text dictation, and a document-chat mode that pulls relevant passages from your local files. For anyone juggling several API keys or keeping models on their own hardware, it folds all of that into one window.