Overview
PromptHouse MCP Server enables your PromptHouse prompt library to be consumed by MCP-enabled AI clients such as Claude Desktop, Cursor, and others. By implementing the Model Context Protocol, it lets applications programmatically discover, retrieve, and execute prompts from your collection, reducing manual copy-paste and enabling automated prompt selection. The server supports web (online PromptHouse service) and local (localhost) connection modes, configurable via environment variables and Claude Desktop settings. You can run it via NPX, install globally, or execute directly from GitHub, and you obtain an access link from your PromptHouse account. Exposed MCP tools include get_prompt_list and get_prompt, for listing prompts with metadata and retrieving content by ID. Advanced configuration includes custom endpoints, a legacy HTTP bridge, and DXT extension for one-click installation, plus debug and timeout controls. It emphasizes privacy with a macOS client that stores data locally (no accounts, no cloud sync) while supporting native model providers and local inference via Ollama. Overall, it’s designed to streamline prompt discovery and execution across MCP-enabled AI tools in productive workflows.
Features
Prompt Management
Save, tag, and manage your prompt library with a clean interface for quick viewing, editing, and filtering by tags.
Prompt Calling
Connect via MCP to allow clients to fetch and execute prompts directly from your collection.
Prompt Recommendations
Access a built-in collection of high-quality prompts for tasks like Productivity and Image Generation.
Privacy-First macOS Client
Native macOS app storing data locally with no accounts or cloud sync; supports major model providers and Ollama local inference.
Who Is This For?
- Prompt engineers:Connect PromptHouse prompts to MCP-enabled clients to fetch and execute prompts automatically.
- AI developers:Expose prompts through MCP for automated retrieval and execution by clients like Claude Desktop.
