thinkinaixyz/deepchat

DeepChat is an open-source desktop application that lets you chat with multiple AI models from a single interface, without switching between different apps or browser tabs. You can connect it to cloud AI services like OpenAI, Anthropic Claude, Google Gemini, DeepSeek, and others, or run local AI models on your own machine through Ollama, which is a tool for running AI models locally without needing a command line. Beyond basic chat, DeepChat includes an agentic layer: it supports MCP (Model Context Protocol), a standard that lets AI models call external tools like running code, browsing the web, or reading files. It also has a Skills system for installing reusable task-specific helpers, and ACP (Agent Client Protocol) support, which lets you treat externally hosted AI agents as if they were just another model to chat with inside the app. The multi-tab, multi-window layout means you can run several AI conversations at the same time without one blocking another. Search is a notable feature: the app can connect to search engines including Google, Bing, Baidu, and others, either through official APIs or by simulating how a browser would visit and read those pages. This lets the AI look up current information as part of answering a question. There is also a remote control feature that lets you trigger DeepChat sessions from messaging platforms including Telegram, Discord, Feishu, and WeChat, useful if you want to query your AI setup from a phone. All conversation data is stored locally by default, and the app supports network proxies for users who need to route traffic through a specific connection. The project is built with TypeScript and released under the Apache 2.0 license, which allows commercial use. Installation involves downloading a pre-built package for Windows, macOS, or Linux from the releases page, then configuring credentials for whichever AI providers you want to use.