drupal-rag-mcp 0.1.0

MCP server for Drupal RAG, file access, routes, and safe Drupal commands

Drupal RAG MCP

A Model Context Protocol (MCP) server that gives AI coding assistants direct access to Drupal projects through:

• Drupal codebase search (RAG)
• File reading
• File writing
• Drupal route discovery
• Safe Drupal command execution
• Project file listing

This server is designed for use with MCP-compatible clients such as:

• Claude Desktop
• Claude Code
• Cursor
• Windsurf
• Cline
• OpenAI Codex CLI
• Other MCP-compatible tools

---

Features

Search Drupal Knowledge Base

Search indexed Drupal source code and documentation using semantic search.

Tool:

search_drupal_docs

Read Drupal Files

Read any file within the configured Drupal project.

Tool:

read_drupal_file

Write Drupal Files

Create or update files inside the Drupal project.

Tool:

write_drupal_file

Run Drupal Commands

Execute approved Drupal development commands.

Supported commands:

• drush
• composer
• phpunit

Tool:

run_drupal_command

List Project Files

Browse project source files without using embeddings.

Tool:

list_project_files

Discover Drupal Routes

Search Drupal routes by:

• Route name
• Path
• Controller
• Form
• Permission
• Title

Tool:

find_routes

---

Architecture

AI Client
    │
    ▼
Drupal MCP Server
    │
    ▼
Drupal RAG API
    │
    ├── Chroma Vector Database
    ├── Drupal Source Code
    └── Drupal Documentation

The MCP server communicates with a local FastAPI service that performs indexing, search, file operations, and command execution.

---

Installation

Install from PyPI:

pip install drupal-rag-mcp

Or install with RAG dependencies:

pip install "drupal-rag-mcp[rag]"

---

Requirements

• Python 3.10+
• Drupal project
• Composer
• Drush (optional)
• ChromaDB

---

Build Your Drupal RAG

Create a .drupal-rag folder under your home directory

mkdir ~/.drupal-rag 
cd ~/.drupal-rag
touch config.toml

Paste this settings

drupal_root = "/path_to_drupal_project_directory"
db_dir = "~/.drupal-rag/chroma"

Inside ~/.drupal-rag

mkdir docs

Inside docs add all your drupal documentation (PDF, MD, txt). Start indexing your docs to create rag

drupal-rag-index

This will:

1. Scan Drupal files
2. Split documents into chunks
3. Generate embeddings
4. Store vectors in Chroma

---

Start the API Server

drupal-rag-api

The API provides:

POST /search
POST /read-file
POST /write-file
POST /run-command
POST /list-files
POST /find-routes

---

Start the MCP Server

drupal-rag-mcp

The MCP server uses stdio transport and is intended to be launched by an MCP client.

---

MCP Client Configuration

Claude Desktop

{
  "mcpServers": {
    "drupal-rag": {
      "command": "drupal-rag-mcp"
    }
  }
}

Cursor

{
  "mcpServers": {
    "drupal-rag": {
      "command": "drupal-rag-mcp"
    }
  }
}

VSCODE

Installing and Running Ollama on Mac

brew install ollama
brew services start ollama
brew services stop ollama

Run and Download an AI Model

ollama pull qwen3
ollama run qwen3

Set Up Continue

1. Add Extension Continue
2. Configure Continue config.yaml

name: Local Config
version: 1.0.0
schema: v1

models:
  - name: Qwen3
    provider: ollama
    model: qwen3

3. Add MCP server in Continue

which drupal-rag-mcp

Example output: /opt/anaconda3/envs/drupal_rag/bin/drupal-rag-mcp

Configure Continue mcp-server.yaml

name: Drupal RAG MCP
version: 0.0.1
schema: v1

mcpServers:
  - name: drupal-rag
    command: /opt/anaconda3/envs/drupal_rag/bin/drupal-rag-mcp
    args: []
    env:
      DRUPAL_ROOT: /path_to_your_drupal

Codex CLI

[mcp_servers.drupal-rag]
command = "drupal-rag-mcp"
args = []

---

Example Prompts

Find the route responsible for user login.

Search Drupal documentation for cache tags.

Show me all custom routes under /admin.

Read web/modules/custom/example/src/Form/ExampleForm.php

Run drush cr

---

Security

The server includes several safeguards:

• Path traversal protection
• Drupal root restriction
• Command allow-list
• Ignored directories and file types
• Local-only API access

Only approved commands can be executed.

Review the source code before enabling write access in production environments.

---

License

MIT License

Copyright (c) 2026 John Paul Mariano

See the LICENSE file for full license text.