Summary
A prompt library the model can edit, not only read from
add_prompt and delete_prompt make the collection writable from inside the conversation, so a prompt refined during a session can be saved without switching to an editor. create_structured_prompt is the difference between a folder of text files and a library — it records author, category, difficulty and tags, which is what makes the listing worth reading months later. Lookup is by name, so a session that cannot recall the name starts with list_prompts.
What it is
A file-backed prompt library with 5 tools: prompts are stored under a filename you choose and retrieved, listed or removed by name.
What you get
- A stored prompt collection the assistant writes into directly, each entry held under its own filename
- Retrieval of a single prompt by name, and a listing of everything currently in the collection
- A guided create path that captures author, category, description, difficulty, tags and title alongside the prompt body
- Removal of a prompt by name once it stops earning its place
Requirements
No account and no key; the collection lives under filenames on whatever machine runs the server.
Setup effort
One command — npm install -g prompts-mcp-server