A high-performance C++ MCP (Model Context Protocol) server for real-time memory scanning and analysis of Windows processes.
Memory MCP provides direct access to process memory through the Model Context Protocol, enabling AI assistants to scan, filter, and analyze memory contents in real-time. Built with modern C++ and optimized for Windows, it offers low-latency memory operations with minimal resource overhead.
- Real-time Memory Scanning: Scan process memory for specific values (strings, integers, doubles)
- Address Filtering: Filter memory addresses based on value changes
- Process Management: Automatic process discovery and memory access
- MCP Protocol: Native MCP server implementation via stdin/stdout
- HTTP API: Alternative HTTP interface for external integrations
- Windows Optimized: Built specifically for Windows with native API calls
- A PC and basic technical knowledge (recommended)
- CMake 3.16+
- Administrator privileges
git clone https://github.com/kkkeeetttx/Memory-MCP
cd Memory-MCPmkdir build
cd build
cmake ..
cmake --build . --config ReleaseThe executable will be located at:
build/exe/memory-mcp-server.exe
Run the server in MCP mode for direct integration with AI tools:
memory-mcp-server.exe --mcpThis mode communicates via stdin/stdout using the MCP protocol, making it ideal for integration with AI assistants like Cursor.
Run the server as an HTTP server for external integrations:
memory-mcp-server.exeThe server will start on port 3000 and provide REST API endpoints.
The server provides the following MCP tools:
Scans process memory for a specific value.
Parameters:
process_name(string): Name of the target processvalue(string): Value to search forvalue_type(string): Type of value ("string", "int", "double")
Returns:
count(integer): Number of addresses foundaddresses(array): List of memory addresses
Retrieves previously found memory addresses.
Parameters:
max_count(integer): Maximum number of addresses to return
Returns:
addresses(array): List of memory addresses
Filters addresses based on value changes.
Parameters:
addresses(array): List of addresses to filternew_value(string): New value to search forvalue_type(string): Type of value
Returns:
filtered_addresses(array): List of addresses that still contain the value
Resets the memory scanner state.
Parameters: None
Returns: Success status
Initialize MCP connection.
List available MCP tools.
Execute MCP tool calls.
Add to your MCP configuration (~/mcp.json):
{
"mcpServers": {
"Memory_MCP": {
"command": "C:/path/to/memory-mcp-server.exe",
"args": ["--mcp"]
}
}
}- Process Isolation: Only scan trusted processes
- Memory Protection: Respect Windows memory protection mechanisms
- Resource Limits: Monitor memory usage during large scans
- Scan Speed: Optimized for real-time scanning of large memory regions
- Memory Usage: Minimal overhead with efficient address tracking
- CPU Usage: Non-blocking operations with configurable scan intervals
-
"Access Denied" Errors
- Ensure running with administrator privileges
- Check if target process is protected
-
Build Failures
- Verify Visual Studio C++ tools are installed
- Ensure CMake version is 3.16 or higher
-
MCP Connection Issues
- Check executable path in MCP configuration
- Verify
--mcpargument is provided
Build with debug configuration for detailed logging:
cmake --build . --config Debug- Dependencies: Header-only libraries (nlohmann/json, httplib)
- Compiler: MSVC with C++17 support
- Build System: CMake 3.16+
- MemoryScanner: Core memory scanning logic
- HttpServer: HTTP API implementation
- MCP Protocol: Native MCP server implementation
This is the first stable release (v1.0.0) of Memory MCP Server.
The project is actively maintained and will receive regular updates with new features, performance improvements, and bug fixes.
- Enhanced Memory Analysis: Pattern recognition and memory structure analysis
- Cross-Platform Support: Linux and macOS compatibility
- Advanced Filtering: Regex-based memory search and multi-value filtering
- Performance Monitoring: Built-in benchmarks and optimization tools
- Plugin System: Extensible architecture for custom memory scanners
- Web Dashboard: Real-time memory monitoring interface
- Memory Dumping: Export memory regions for external analysis
- Process Injection Detection: Security-focused memory scanning features
- v1.0.0 (Current): Core MCP server with basic memory scanning capabilities
- v1.1.0 (Planned): Enhanced filtering and performance optimizations
- v1.2.0 (Planned): Advanced memory analysis tools
- v1.3.0 (Planned): Python/NodeJS integration
- v2.0.0 (Future): Major architecture improvements
Contributions are welcome! Whether it's bug reports, feature requests, code improvements, or documentation updates - every contribution helps make Memory MCP Server better.