Crush innovatively proposes the Model Context Protocol (MCP) to enable feature extensions via transports such as HTTP, stdio or SSE. Developers can configure remote MCP endpoints to connect to external services or integrate local toolchains into AI workflows. This protocol design gives Crush the possibility to scale beyond a normal AI assistant, theoretically interfacing with any compatible third-party service.
Typical application scenarios include connecting to internal knowledge base systems, calling internal API documentation services, or integrating proprietary code quality inspection tools.The open nature of the MCP protocol allows Crush to adapt to enterprise-level development environments and meet customization needs.
This answer comes from the articleCrush: endpoint AI programming assistant with integrated LSP and multi-model switchingThe
English 
简体中文 
日本語 
Deutsch 
Português do Brasil