Reading somebody's published system prompt teaches you something; loading it as the working instructions teaches you more. Because every file in the directory becomes a tool automatically, the same server doubles as the place to keep your own prompts.
A server that exposes a collection of published system prompts, summaries and tool definitions from well-known AI tools. Every file in its prompts directory becomes a tool of its own — named service-variant-flavour, such as cursor-agent-system or devin-summary — alongside two calls for browsing and picking one. Each call returns the prompt with a reminder for the model to adopt it.
- The available prompts browsed, filtered by service, flavour or provider
- Ranked suggestions for your LLM and your keywords, so the pick is matched to the model you are on
- Direct access to any individual prompt as its own tool
- Automatic discovery: .txt, .md, .yaml, .yml and .json files dropped into the prompts directory become tools, with service, variant and LLM family inferred
- Your own prompts added the same way, without touching the server code
A recent Node.js. The README's npm name, system-prompts-mcp-server, is not on the registry; the repository also offers a one-click install link for Cursor. No account and no key — the prompts ship with the repository.