Switchbox
Send one message to up to four models at once and watch the replies stream side by side, each pane with its own model, system prompt and settings.
A local web app for putting the same question to several LLMs. OpenRouter, OpenAI, Anthropic, Ollama, LM Studio and any OpenAI-compatible endpoint sit behind one composer, and a comparison view shows each reply's speed, tokens and cost, with a word-level diff between any two panes.
Models get tools: web search and page fetching, a JavaScript sandbox, long-term memory, MCP servers, and a per-chat workspace folder they can read, write and, if you allow it, run commands in. Every tool is Auto, Ask or Off. History is stored in SQLite through repolayer, and the server binds to 127.0.0.1 and refuses requests that are not addressed to localhost.