A Desktop Control Panel for LLM Tool Chaos

What it does Dive is a desktop application that acts as a host for Model Context Protocol (MCP) servers, connecting them to LLMs that support function calling. It works with OpenAI, Anthropic, Ollama, and compatible models, giving the LLM access to external tools like file management, bash execution, and web fetching. The app handles both local stdio and SSE connections, plus managed cloud servers through an integrated hub.

The interesting bit The developers clearly recognize that MCP onboarding is painful: they ship an “MCP Server Installer Agent” that automates server setup, and offer one-click cloud-managed servers via OAPHub.ai for users who would rather not wrestle with local Python and Node.js dependencies. The project also builds in both Electron and Tauri flavors, letting users trade stability for a sub-30 MB installer on Windows and Linux.

Key highlights

• Supports local MCP servers over stdio and SSE, or managed cloud instances via OAPHub.ai
• Built-in local tools for Fetch, File Manager, and Bash execution available out of the box
• Granular toggling of individual MCP tools per server for precise control
• Dual architecture offering both a traditional Electron build and a lighter Tauri build
• 24+ language localizations, custom system prompts, and chat draft auto-saving

Caveats

• MCP server authentication is explicitly marked unstable and may require frequent re-authorization
• The Tauri build is not yet available for macOS
• macOS and Linux users must supply their own Python and Node.js environments; they are not bundled

Verdict Developers already experimenting with MCP servers and looking for a graphical control panel will find Dive a practical step up from terminal-only workflows. Those expecting a fully zero-config, cross-platform desktop experience should note the missing macOS Tauri build and manual environment requirements on some platforms.