Dive: Open source MCP host desktop application

The OpenAgentPlatform-Dive integrates seamlessly with any LLM that supports function invocation:

Project introduction

Dive Is an open source desktop application positioned as MCP Host(That is, the desktop client hosting MCP Server), the main features are:

Supports multiple LLM (Large Language Models): Compatible with ChatGPT, Anthropic Claude, Ollama, local models, etc., provided that the model has the ability to function calling.
cross-platform: Provides native installation packages for Windows, macOS and Linux.
Using Model Context Protocol (MCP): Support stdio and SSE modes for easy connection to MCP servers (local or remote).
Language diversity interface: Support Traditional Chinese, Simplified Chinese, English, Spanish, Japanese, Korean, etc.
feature-rich: Includes multi-API-key management, customized system prompts, MCP tools configured in the UI, automatic update mechanisms, etc.

In fact, starting from v0.8.0, MCP Host has been rewritten from TypeScript to Python and split into independent projects. dive-mcp-host：

Provide unified API interface: Supports functions such as managing multiple LLMs, session persistence, HTTP Websocket services, multi-threaded conversations, and user management.
Use with Dive Desktop: The desktop is responsible for task presentation and tool invocation, and MCP Host provides back-end functional support.
development environment: Applies to Python ≥ 3.12, supports command line tools and HTTP services.

According to the official update log and Reddit community feedback:

Launch the Python version of MCP Host (dive-mcp-host) to better support the popular LangChain in the Python ecosystem
Enhance the LLM setting interface: You can edit API keys, customize model IDs, and choose whether to verify whether the model supports tool calling
Enhance tool management: You can directly add, edit, and delete MCP tools within the UI, and support JSON editing and form mode.
Although Python refactoring briefly interrupted development, it laid the foundation for subsequent stronger functions.

Installation and download: Go to the Release page and select the corresponding platform package (.exe/.dmg/.AppImage）
- Add multiple LLM API keys and model IDs to the GUI.
- Configure the MCP server, connect to local (Python backend) or remote SSE type servers, and add tools (such as fetch, yt‑dl, etc.)
- Using Python projects dive-mcp-host Start an HTTP service or command-line interaction

The community considers it a “Lightweight, simple, truly usable local LLM + function tool integrated solution”
Recommended as a standard MCP Client example on Hacker News and registration platforms such as PulseMCP.

ability	described
Multiple model support	Any LLM that supports function calling
Desktop back-end separation	Front-end GUI + Python backend
Tool chain is open	Support custom tool plug-ins
Active community	GitHubˇ 1.4k, Fork107, Reddit and developers continue to follow up

It is suitable for developers and advanced users who want to quickly build a local LLM + tool calling system and want a GUI operating experience.

trial experience: Install the latest Release, connect to the local or remote MCP Server, and invoke a simple tool (such as fetch).
development and debugging: Clone project, overwrite the tool configuration, or try to connect to another LLM using the Python version of MCP Host.
participate in community: Submit issues, PR, or go to Discord/Reddit to communicate with developers.

Oil tubing: