Cursor DB MCP Server
by jbdamask
import os
import json
import sqlite3
import platform
import re
from pathlib import Path
import argparse
import logging
from typing import Dict, List, Optional, Any, Union, AsyncIterator
from contextlib import asynccontextmanager
import sys
# Configure logging
logging.basicConfig(
level=logging.INFO,
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
handlers=[
logging.FileHandler(os.path.join(os.path.dirname(os.path.abspath(__file__)), "mcp-server.log")),
logging.StreamHandler()
]
)
logger = logging.getLogger('cursor-mcp')
# Import MCP libraries
try:
from mcp.server.fastmcp import FastMCP, Context
except ImportError as e:
logger.error(f"Failed to import MCP libraries: {str(e)}. Make sure they are installed.")
sys.exit(1)
# Global DB manager instance
db_manager = None
class CursorDBManager:
def __init__(self, cursor_path=None, project_dirs=None):
"""
Initialize the CursorDBManager with a Cursor main directory and/or list of project directories.
Args:
cursor_path (str): Path to main Cursor directory (e.g. ~/Library/Application Support/Cursor/User/)
project_dirs (list): List of paths to Cursor project directories containing state.vscdb files
"""
if cursor_path:
self.cursor_path = Path(cursor_path).expanduser().resolve()
else:
# Try to get the default cursor path
self.cursor_path = self.get_default_cursor_path()
self.project_dirs = project_dirs or []
self.db_paths = {}
self.projects_info = {}
self.global_db_path = None
self.refresh_db_paths()
def get_default_cursor_path(self):
"""Return the default Cursor path based on the operating system"""
system = platform.system()
home = Path.home()
default_path = None
if system == "Darwin": # macOS
default_path = home / "Library/Application Support/Cursor/User"
elif system == "Windows":
default_path = home / "AppData/Roaming/Cursor/User"
elif system == "Linux":
default_path = home / ".config/Cursor/User"
else:
logger.warning(f"Unknown operating system: {system}. Cannot determine default Cursor path.")
return None
logger.info(f"Detected default Cursor path for {system}: {default_path}")
return default_path
def detect_cursor_projects(self):
"""Detect Cursor projects by scanning the workspaceStorage directory"""
if not self.cursor_path:
logger.error("No Cursor path available")
return []
# Check if the path exists
if not self.cursor_path.exists():
logger.error(f"Cursor path does not exist: {self.cursor_path}")
return []
workspace_storage = self.cursor_path / "workspaceStorage"
if not workspace_storage.exists():
logger.warning(f"Workspace storage directory not found: {workspace_storage}")
return []
logger.info(f"Found workspace storage directory: {workspace_storage}")
projects = []
# Scan all subdirectories in workspaceStorage
for workspace_dir in workspace_storage.iterdir():
if not workspace_dir.is_dir():
continue
workspace_json = workspace_dir / "workspace.json"
state_db = workspace_dir / "state.vscdb"
if workspace_json.exists() and state_db.exists():
try:
with open(workspace_json, 'r') as f:
workspace_data = json.load(f)
folder_uri = workspace_data.get("folder")
if folder_uri:
# Extract the project name from the URI
# For "file:///Users/johndamask/code/cursor-chat-browser", get "cursor-chat-browser"
project_name = folder_uri.rstrip('/').split('/')[-1]
projects.append({
"name": project_name,
"db_path": str(state_db),
"workspace_dir": str(workspace_dir),
"folder_uri": folder_uri
})
logger.info(f"Found project: {project_name} at {state_db}")
except Exception as e:
logger.error(f"Error processing workspace: {workspace_dir}: {e}")
return projects
def refresh_db_paths(self):
"""Scan project directories and identify all state.vscdb files"""
self.db_paths = {}
self.projects_info = {}
# First, detect projects from the Cursor directory
if self.cursor_path:
cursor_projects = self.detect_cursor_projects()
for project in cursor_projects:
project_name = project["name"]
self.db_paths[project_name] = project["db_path"]
self.projects_info[project_name] = project
# Set the global storage database path
global_storage_path = self.cursor_path / "globalStorage" / "state.vscdb"
if global_storage_path.exists():
self.global_db_path = str(global_storage_path)
logger.info(f"Found global storage database at {self.global_db_path}")
else:
logger.warning(f"Global storage database not found at {global_storage_path}")
# Then add explicitly specified project directories
for project_dir in self.project_dirs:
project_path = Path(project_dir).expanduser().resolve()
db_path = project_path / "state.vscdb"
if db_path.exists():
project_name = project_path.name
self.db_paths[project_name] = str(db_path)
self.projects_info[project_name] = {
"name": project_name,
"db_path": str(db_path),
"workspace_dir": None,
"folder_uri": None
}
logger.info(f"Found database: {project_name} at {db_path}")
else:
logger.warning(f"No state.vscdb found in {project_path}")
# def add_project_dir(self, project_dir):
# """Add a new project directory to the manager"""
# project_path = Path(project_dir).expanduser().resolve()
# if project_path not in self.project_dirs:
# self.project_dirs.append(project_path)
# self.refresh_db_paths()
# return len(self.db_paths)
def list_projects(self, detailed=False):
"""
Return list of available projects
Args:
detailed (bool): Whether to return detailed project information
Returns:
dict: Project information (either just DB paths or full details)
"""
if detailed:
return self.projects_info
return self.db_paths
def execute_query(self, project_name, table_name, query_type, key=None, limit=100):
"""
Execute a query against a specific project's database
Args:
project_name (str): Name of the project (key in db_paths)
table_name (str): Either 'ItemTable' or 'cursorDiskKV'
query_type (str): Type of query ('get_all', 'get_by_key', 'search_keys')
key (str, optional): Key to search for when using 'get_by_key' or 'search_keys'
limit (int): Maximum number of results to return
Returns:
list: Query results
"""
if project_name not in self.db_paths:
raise ValueError(f"Project '{project_name}' not found")
if table_name not in ["ItemTable", "cursorDiskKV"]:
raise ValueError("Table name must be either 'ItemTable' or 'cursorDiskKV'")
db_path = self.db_paths[project_name]
try:
conn = sqlite3.connect(db_path)
cursor = conn.cursor()
if query_type == "get_all":
cursor.execute(f"SELECT key, value FROM {table_name} LIMIT ?", (limit,))
elif query_type == "get_by_key" and key:
cursor.execute(f"SELECT key, value FROM {table_name} WHERE key = ?", (key,))
elif query_type == "search_keys" and key:
search_term = f"%{key}%"
cursor.execute(f"SELECT key, value FROM {table_name} WHERE key LIKE ? LIMIT ?",
(search_term, limit))
else:
raise ValueError("Invalid query type or missing key parameter")
results = []
for row in cursor.fetchall():
key, value = row
try:
# Try to parse JSON value
parsed_value = json.loads(value)
results.append({"key": key, "value": parsed_value})
except json.JSONDecodeError:
# If not valid JSON, return as string
results.append({"key": key, "value": value})
conn.close()
return results
except sqlite3.Error as e:
logger.error(f"SQLite error: {e}")
raise
def get_chat_data(self, project_name):
"""
Retrieve AI chat data from a project
Args:
project_name (str): Name of the project
Returns:
dict: Chat data from the project
"""
if project_name not in self.db_paths:
raise ValueError(f"Project '{project_name}' not found")
try:
results = self.execute_query(
project_name,
"ItemTable",
"get_by_key",
"workbench.panel.aichat.view.aichat.chatdata"
)
if results and len(results) > 0:
return results[0]["value"]
else:
return {"error": "No chat data found for this project"}
except Exception as e:
logger.error(f"Error retrieving chat data: {e}")
raise
def get_composer_ids(self, project_name):
"""
Retrieve composer IDs from a project
Args:
project_name (str): Name of the project
Returns:
list: List of composer IDs
"""
if project_name not in self.db_paths:
raise ValueError(f"Project '{project_name}' not found")
try:
results = self.execute_query(
project_name,
"ItemTable",
"get_by_key",
"composer.composerData"
)
if results and len(results) > 0:
composer_data = results[0]["value"]
# Extract composer IDs from the data
composer_ids = []
if "allComposers" in composer_data:
for composer in composer_data["allComposers"]:
if "composerId" in composer:
composer_ids.append(composer["composerId"])
return {
"composer_ids": composer_ids,
"full_data": composer_data
}
else:
return {"error": "No composer data found for this project"}
except Exception as e:
logger.error(f"Error retrieving composer IDs: {e}")
raise
def get_composer_data(self, composer_id):
"""
Retrieve composer data from global storage
Args:
composer_id (str): Composer ID
Returns:
dict: Composer data
"""
if not self.global_db_path:
raise ValueError("Global storage database not found")
try:
conn = sqlite3.connect(self.global_db_path)
cursor = conn.cursor()
key = f"composerData:{composer_id}"
cursor.execute("SELECT value FROM cursorDiskKV WHERE key = ?", (key,))
row = cursor.fetchone()
conn.close()
if row:
try:
return {"composer_id": composer_id, "data": json.loads(row[0])}
except json.JSONDecodeError:
return {"composer_id": composer_id, "data": row[0]}
else:
return {"error": f"No data found for composer ID: {composer_id}"}
except sqlite3.Error as e:
logger.error(f"SQLite error: {e}")
raise
# Create an MCP server with lifespan support
@asynccontextmanager
async def app_lifespan(app: FastMCP) -> AsyncIterator[Dict[str, Any]]:
"""Manage application lifecycle with context"""
try:
# Initialize the DB manager on startup
global db_manager
db_manager = CursorDBManager()
# Parse command line arguments
parser = argparse.ArgumentParser(description='Cursor IDE SQLite Database MCP Server')
parser.add_argument('--cursor-path', help='Path to Cursor User directory (e.g. ~/Library/Application Support/Cursor/User/)')
parser.add_argument('--project-dirs', nargs='+', help='List of additional Cursor project directories to scan')
# Parse known args only, to avoid conflicts with MCP's own args
args, _ = parser.parse_known_args()
# Configure the DB manager with the Cursor path
if args.cursor_path:
db_manager.cursor_path = Path(args.cursor_path).expanduser().resolve()
# Add explicitly specified project directories
if args.project_dirs:
for project_dir in args.project_dirs:
db_manager.add_project_dir(project_dir)
# Log detected Cursor path
if db_manager.cursor_path:
logger.info(f"Using Cursor path: {db_manager.cursor_path}")
else:
logger.warning("No Cursor path specified or detected")
logger.info(f"Available projects: {list(db_manager.list_projects().keys())}")
# Yield empty context - we're using global db_manager instead
yield {}
finally:
# Cleanup on shutdown (if needed)
logger.info("Shutting down Cursor DB MCP server")
# Create the MCP server with lifespan
mcp = FastMCP("Cursor DB Manager", lifespan=app_lifespan)
# MCP Resources
@mcp.resource("cursor://projects")
def list_all_projects() -> Dict[str, str]:
"""List all available Cursor projects and their database paths"""
global db_manager
return db_manager.list_projects(detailed=False)
@mcp.resource("cursor://projects/detailed")
def list_detailed_projects() -> Dict[str, Dict[str, Any]]:
"""List all available Cursor projects with detailed information"""
global db_manager
return db_manager.list_projects(detailed=True)
@mcp.resource("cursor://projects/{project_name}/chat")
def get_project_chat_data(project_name: str) -> Dict[str, Any]:
"""Retrieve AI chat data from a specific Cursor project"""
global db_manager
try:
return db_manager.get_chat_data(project_name)
except ValueError as e:
return {"error": str(e)}
except Exception as e:
return {"error": f"Error retrieving chat data: {str(e)}"}
@mcp.resource("cursor://projects/{project_name}/composers")
def get_project_composer_ids(project_name: str) -> Dict[str, Any]:
"""Retrieve composer IDs from a specific Cursor project"""
global db_manager
try:
return db_manager.get_composer_ids(project_name)
except ValueError as e:
return {"error": str(e)}
except Exception as e:
return {"error": f"Error retrieving composer data: {str(e)}"}
@mcp.resource("cursor://composers/{composer_id}")
def get_composer_data_resource(composer_id: str) -> Dict[str, Any]:
"""Retrieve composer data from global storage"""
global db_manager
try:
return db_manager.get_composer_data(composer_id)
except ValueError as e:
return {"error": str(e)}
except Exception as e:
return {"error": f"Error retrieving composer data: {str(e)}"}
# MCP Tools
@mcp.tool()
def query_table(project_name: str, table_name: str, query_type: str, key: Optional[str] = None, limit: int = 100) -> List[Dict[str, Any]]:
"""
Query a specific table in a project's database
Args:
project_name: Name of the project
table_name: Either 'ItemTable' or 'cursorDiskKV'
query_type: Type of query ('get_all', 'get_by_key', 'search_keys')
key: Key to search for when using 'get_by_key' or 'search_keys'
limit: Maximum number of results to return
Returns:
List of query results
"""
global db_manager
try:
return db_manager.execute_query(project_name, table_name, query_type, key, limit)
except ValueError as e:
return [{"error": str(e)}]
except sqlite3.Error as e:
return [{"error": f"Database error: {str(e)}"}]
@mcp.tool()
def refresh_databases() -> Dict[str, Any]:
"""Refresh the list of database paths"""
global db_manager
db_manager.refresh_db_paths()
return {
"message": "Database paths refreshed",
"projects": db_manager.list_projects()
}
# @mcp.tool()
# def add_project_directory(project_dir: str) -> Dict[str, Any]:
# """
# Add a new project directory to the manager
# Args:
# project_dir: Path to the project directory
# Returns:
# Result of the operation
# """
# global db_manager
# try:
# count = db_manager.add_project_dir(project_dir)
# return {
# "message": f"Project directory added. Total projects: {count}",
# "projects": db_manager.list_projects()
# }
# except Exception as e:
# return {"error": f"Error adding project directory: {str(e)}"}
# MCP Prompts
@mcp.prompt()
def explore_cursor_projects() -> str:
"""Create a prompt to explore Cursor projects"""
return """
I can help you explore your Cursor projects and their data.
Here are some things I can do:
1. List all your Cursor projects
2. Show AI chat history from a project
3. Find composer data
4. Query specific tables in the Cursor database
What would you like to explore?
"""
@mcp.prompt()
def analyze_chat_data(project_name: str) -> str:
"""Create a prompt to analyze chat data from a specific project"""
return f"""
I'll analyze the AI chat data from your '{project_name}' project.
I can help you understand:
- The conversation history
- Code snippets shared in the chat
- Common themes or questions
Would you like me to focus on any specific aspect of the chat data?
"""