prompt-library

A collection of AI prompts organised as executable MCP (Model Context Protocol) prompt servers. Each server exposes structured prompts that can be invoked directly in Copilot Chat agent mode using @server-name /prompt-name.

Repository Structure

prompt-library/
│
├── pyproject.toml                      # Deployable Python project config
├── mcp.example.jsonc                   # Example MCP registrations (copy to your MCP config)
├── docs/                               # Usage guides for each server
│   ├── code-prompts-usage.md
│   ├── dev-prompts-usage.md
│   ├── food-prompts-usage.md
│   ├── prompt-template-usage.md
│   └── research-prompts-usage.md
├── src/
│   └── prompt_library/                 # Installable package (source of truth)
│       └── servers/                    # MCP prompt server implementations
│           ├── code_prompts.py         # code-docs-updater, code-docs-creator
│           ├── food_prompts.py         # recipe-generator, ingredient-subs, meal-planner
│           ├── research_prompts.py     # deep-dive, article-summary, competitive-analysis
│           └── dev_prompts.py          # create-mcp-server, add-mcp-prompt
│
├── FrequentlyAskedQuestions.md         # Common questions about MCP server development
├── LICENSE
└── README.md                           # This file

MCP Servers

Quick Start

Install & run (deployable project)

This repository is set up as an installable Python project (via pyproject.toml). With uv:

uv sync
uv run code-prompts

Other console scripts:

• food-prompts
• research-prompts
• dev-prompts

Registration (mcp.json)

This repo ships an example config at mcp.example.jsonc. Copy its entries into the mcp.json location used by your MCP client (VS Code / Copilot, Claude Desktop, etc.).

{
  "code-prompts": {
    "command": "code-prompts",
    "args": []
  }
}

For a no-install alternative, mcp.example.jsonc also includes a uv run --project <path> variant.

Using an existing prompt

All servers must be registered in mcp.json before use. Once registered, invoke any prompt in Copilot Chat:

@food-prompts /recipe-generator ingredients=chicken, garlic, lemon

@research-prompts /deep-dive topic=Model Context Protocol

@research-prompts /competitive-analysis subject="GitHub Copilot" competitor=Cursor

Building a new server

The dev-prompts server provides two prompts for scaffolding new servers without writing code manually:

Step 1 — Generate the server file:

@dev-prompts /create-mcp-server
  server_name=my-server
  prompt_name=my-prompt
  prompt_description="what the prompt does"
  arguments="arg_one (description), arg_two (description), arg_three (description)"
  persona=expert persona
  output_file=my-server.py

Step 2 — Add more prompts (optional):

@dev-prompts /add-mcp-prompt
  prompt_name=my-second-prompt
  description="what this prompt does"
  arguments="required_arg* (description), optional_arg? (description)"
  server_file=my-server.py

Step 3 — Create the usage documentation:

@code-prompts /code-docs-creator code_file_to_read=src/prompt_library/servers/my_server.py

Step 4 — Keep documentation up to date:

@code-prompts /code-docs-updater
  code_file_to_read=src/prompt_library/servers/my_server.py
  file_to_update=docs/my-server-usage.md

See the usage guides in docs/ for the current prompt and transport details.

Hosted (Render) streamable-http

Use this when you want a single hosted deployment per code update and your MCP client can connect over HTTP (streamable-http).

Start command (Render)

Render provides $PORT. Bind to all interfaces so Render can route traffic:

code-prompts --transport streamable-http --host 0.0.0.0 --port $PORT --path /mcp

The other servers work the same way:

food-prompts --transport streamable-http --host 0.0.0.0 --port $PORT --path /mcp
research-prompts --transport streamable-http --host 0.0.0.0 --port $PORT --path /mcp
dev-prompts --transport streamable-http --host 0.0.0.0 --port $PORT --path /mcp

Notes

• The HTTP endpoint is /mcp by default (override with --path or $MCP_HTTP_PATH).
• Add authentication if you deploy to the public internet.
• See the server usage guides in docs/ for details.

Documentation