A desktop AI client that treats agents as just another model

What it does DeepChat is an Electron-based desktop client that talks to cloud LLMs (OpenAI, Gemini, Anthropic, DeepSeek, and others) and local Ollama models in one interface. It wraps chat, tool calling, and agent runtimes into a multi-tab, multi-window app that also lets you control sessions remotely from Telegram, Discord, Feishu, QQBot, or WeChat iLink.

The interesting bit Instead of treating external agents as plugins, DeepChat uses the Agent Client Protocol (ACP) to surface them as selectable models in the model picker, complete with a dedicated workspace UI for plans and terminal output. It also implements the Model Context Protocol (MCP) for tool calling and supports installable Skills that load task-specific instructions per conversation.

Key highlights

• Supports a wide range of cloud providers and local Ollama deployment without leaving the GUI
• Built-in MCP client with Resources, Prompts, and Tools support, plus inMemory services for code execution and file operations
• Installable Skills via folder, ZIP, or URL, compatible with other AI coding assistants like Claude Code and Cursor
• Remote session control via Telegram, Discord, Feishu/Lark, QQBot, and WeChat iLink
• Multi-window, multi-tab architecture with conversation forking, message retry/variations, and Artifacts rendering

Verdict Worth a look if you want a single desktop hub for both commercial and local LLMs with protocol-level tool and agent support. Skip it if you already have a terminal-based workflow you love or if you dread Electron apps on principle.