promptbook

Personal cookbook for AI prompts - MCP Server with RAG-powered semantic search

Installation

Run this command in your terminal to add the MCP server to Claude Code.

Run in terminal:

Command

claude mcp add --transport stdio isaacpalomero-promptbook-mcp python mcp_server.py \
  --env LOG_LEVEL="Python logging level (default: INFO)" \
  --env CHUNK_SIZE="Prompt chunk size (default: 500)" \
  --env ENABLE_RAG="Toggle RAG initialization (default: true)" \
  --env PROMPTS_DIR="Root folder for categorized prompts (default: ./prompts)" \
  --env LMSTUDIO_URL="LMStudio endpoint URL (default: http://localhost:1234)" \
  --env SESSIONS_DIR="Directory watched for exported sessions (default: ./sessions)" \
  --env CHUNK_OVERLAP="Prompt chunk overlap (default: 100)" \
  --env VECTOR_DB_DIR="Persistent ChromaDB path (default: <PROMPTS_DIR>/.vectordb)" \
  --env LMSTUDIO_MODEL="LMStudio embedding model name (default: nomic-embed-text)" \
  --env EMBEDDING_MODEL="Embedding model name for sentence-transformers (default: all-MiniLM-L6-v2)" \
  --env EMBEDDING_PROVIDER="Embedding backend ('sentence-transformer' or 'lmstudio')" \
  --env LMSTUDIO_DIMENSION="LMStudio embedding size (default: 768)" \
  --env AUTO_REINDEX_INTERVAL="Seconds between auto-index checks (default: 30)"

How to use

Promptbook MCP is a Python-based server that automatically captures prompts from your AI sessions, indexes them with a semantic search (RAG), and exposes a set of MCP Tools to manage and query your prompt library. Once running, you can connect via any MCP client (including Claude Desktop or custom MCP clients) and use tools like search_prompts to semantically locate prompts, create_prompt to add new entries, update_prompt and delete_prompt to maintain your collection, and index_prompts to rebuild the search index. The server also supports organizing sessions, retrieving full prompt content, and browsing by category, making it easy to curate a structured and searchable knowledge base of coding prompts. Typical workflows involve saving prompts from AI-assisted sessions, indexing them for fast search, and then reusing or refining prompts for future tasks across development scenarios such as refactoring, testing, debugging, and code reviews.

How to install

{"Prerequisites:","- Python 3.9+ (or Docker for containerized setup)","- Git","- Optional: Docker installed if using docker-compose method","","Automated setup (recommended):","bash","git clone https://github.com/isaacpalomero/promptbook-mcp.git","cd promptbook-mcp","./setup.sh","","","Manual Python setup (local):","bash","# Clone repository","git clone https://github.com/isaacpalomero/promptbook-mcp.git","cd promptbook-mcp","","# Create and activate virtual environment","python3 -m venv .venv","source .venv/bin/activate # Windows: .venv\\Scripts\\activate","","# Install dependencies","pip install -r requirements.txt","","# Optionally copy example config","cp .env.example .env","","# Start server","python mcp_server.py","","","Docker method:","bash","git clone https://github.com/isaacpalomero/promptbook-mcp.git","cd promptbook-mcp","","# Copy environment file and start services","cp .env.example .env","docker-compose up -d","","# Verify logs","docker-compose logs",""}

Additional notes

Tips and considerations:

The server exposes a rich MCP Tools interface; you can perform semantic search with search_prompts and manage prompts with create_prompt, update_prompt, and delete_prompt.
Embedding provider is configurable via EMBEDDING_PROVIDER (sentence-transformer or lmstudio). If using lmstudio, ensure LMSTUDIO_URL and related settings are correct.
RAG indexing can run automatically based on AUTO_REINDEX_INTERVAL; you can also explicitly run index_prompts via MCP clients.
Environment variables are centralized in the .env file or CONFIG; you can customize prompts_dir, sessions_dir, and vector DB location to fit your environment.
For local testing, ensure you have at least 2GB RAM and Python 3.9+; Docker is supported for containerized deployment and easier isolation.

Related MCP Servers

Pare

Dev tools, optimized for agents. Structured, token-efficient MCP servers for git, test runners, npm, Docker, and more.

unity

A Unity MCP server that allows MCP clients like Claude Desktop or Cursor to perform Unity Editor actions.

local -gateway

Aggregate multiple MCP servers into a single endpoint with web UI, OAuth 2.1, and profile-based tool management

memory

A MCP (Model Context Protocol) server providing long-term memory for LLMs

shodan

Shodan MCP server for Claude, Cursor & VS Code. 20 tools for passive reconnaissance, CVE/CPE intelligence, DNS analysis, and device search. 4 tools work free without an API key. OSINT and vulnerability research from your IDE.

vscode-context

MCP Server to Connect with VS Code IDE