PyMemoryEditor

What the PyMemoryEditor MCP server does

PyMemoryEditor MCP server exposes the process-memory functions of the PyMemoryEditor Python library through the Model Context Protocol. It allows an MCP-capable assistant to work with a running process rather than requiring every scan operation to be written manually in Python.

The server covers the main Cheat Engine-style workflow: identify a value in memory, narrow the results after the process changes, inspect a matching address, and write a replacement value. The README states that fourteen tools cover the complete workflow. It also supports reading and modifying process memory and working with scan results maintained by the server.

Typical uses include investigating a live program, building game modding or speedrunning utilities, sampling process state for observability, and conducting authorized reverse-engineering research. The underlying library supports Windows, Linux, and macOS, including 32-bit and 64-bit systems.

How it works

An assistant sends MCP tool requests to the local server. The server delegates those operations to PyMemoryEditor, which accesses another process through operating-system APIs. A scan can produce many matches; instead of returning every address to the model, the server stores the results and returns a handle that later requests can use for refinement.

This design supports an iterative workflow. For example, an assistant can search for a value such as a current health amount, ask the user to change the value in the running program, and then refine the existing matches. The resulting address can then be inspected or used in a read or write operation.

The PyMemoryEditor MCP server is a stdio server. Clients start the command locally and communicate with it over the MCP process connection; the README does not describe a hosted endpoint or network deployment mode.

Setup and configuration

Install the MCP extra with:

pip install "PyMemoryEditor[mcp]"

The installed command is pymemoryeditor-mcp. Claude Desktop can be configured with an mcpServers entry whose command is pymemoryeditor-mcp and whose arguments array is empty. The README also shows the equivalent Claude command-line registration:

claude mcp add pymemoryeditor -- pymemoryeditor-mcp

No environment variables or additional configuration values are documented in the supplied material. The package requires Python 3.10 or newer. The optional app extra installs a separate GUI application, while the optional speed extra adds NumPy-based scan acceleration; neither is required for the MCP server.

Tools and capabilities

The documented capabilities include:

• Searching process memory for values.
• Refining previously stored scan results.
• Reading values from a process.
• Writing values to a process.
• Keeping scan results server-side behind a handle.
• Supporting the full assistant-guided scan workflow through fourteen MCP tools.

The underlying project documentation also describes value ranges, byte-pattern and regular-expression scans, pointer chains, live remote pointers, and pointer scans. The supplied README does not map those capabilities to individual MCP tool names, so integrations should rely on the published MCP guide for exact tool schemas and parameters.

Limitations and notes

Access depends on the operating system’s permissions and process-access rules. The project includes separate platform notes and troubleshooting guidance for Windows, Linux, and macOS, but the supplied material does not specify the exact permissions required in each case.

Process-memory access can affect the target program. Use the PyMemoryEditor MCP server only with processes you own or are explicitly authorized to inspect. The README presents game modding, debugging, observability, learning, and authorized security research as possible application areas; it does not establish permission to modify third-party software.

The server is intended for local MCP execution. No environment-variable configuration, remote transport, authentication layer, or client-specific behavior beyond the shown MCP configuration is documented here.