I got tired of typing the same detailed instructions to Claude every time I wanted specific help. "Remember my coding standards, avoid these patterns, format output like this..." over and over.

So I built a simple MCP server that lets you save prompt templates and reuse them with a tag system. Create a prompt once, then just type ">> code-review" or ">> music-discovery" and it pulls in all your detailed instructions with variable substitution.

Under the hood it's just markdown files with {{variable}} syntax. The MCP integration means it works with any client that supports the protocol (I use it with Claude Desktop and Cursor).

The use case that sold me on building this was music recommendations. Instead of re-explaining my Spotify workflow and preferences every time, I created one comprehensive template. Now I just tag it and fill in my current mood. I also use it a ton at work to make progress on Linear tickets or add new APIs.

It's basically solving the "I wish I could save this prompt" problem that comes up when you start using LLMs for regular workflows. Nothing fancy, just practical (ymmv).

Just published, but been using it locally for some time. Hope it proves useful to someone. Feedback welcome.

GitHub: https://github.com/raisinbread/teleprompter NPM: https://www.npmjs.com/package/mcp-teleprompter