cheatengine-mcp-bridge

That's exactly what this does.

- Stop clicking through hex dumps and start having conversations with the memory.

What You Get:

Your AI can now:

• Read any memory instantly (integers, floats, strings, pointers)
• Follow pointer chains: [[base+0x10]+0x20]+0x8 → resolved in ms
• Auto-analyze structures with field types and values
• Identify C++ objects via RTTI: "This is a CPlayer object"
• Disassemble and analyze functions
• Debug invisibly with hardware breakpoints + Ring -1 hypervisor
• And much more!

How It Works

flowchart TD
    AI[AI Agent: Claude/Cursor/Copilot]
    
    AI -->|MCP Protocol - JSON-RPC over stdio| MCP
    
    MCP[mcp_cheatengine.py - Python MCP Server]
    
    MCP <-->|Named Pipe - Async| PIPE
    
    PIPE["\\.\\pipe\\CE_MCP_Bridge_v99"]
    
    PIPE <--> CE
    
    subgraph CE[Cheat Engine - DBVM Mode]
        subgraph LUA[ce_mcp_bridge.lua]
            WORKER[Worker Thread - Blocking I/O]
            MAIN[Main Thread - GUI + CE API]
            WORKER <-->|Sync| MAIN
        end
    end
    
    MAIN -->|Memory Access| TARGET[Target .exe]

Installation

pip install -r MCP_Server/requirements.txt

Or manually:

pip install mcp pywin32

[!NOTE] Native pipe mode is Windows only because it uses Named Pipes (pywin32). Use TCP relay transport when the MCP server runs outside the Windows environment that hosts Cheat Engine and cannot open the named pipe directly.

Quick Start

1. Load Bridge in Cheat Engine

1. Enable DBVM in Cheat Engine if you plan to use DBVM tools.
2. Open Cheat Engine's Lua Engine or script executor.
   • Preferred: File -> Execute Script -> open MCP_Server/ce_mcp_bridge.lua -> Execute.
   • If your Cheat Engine build does not show File -> Execute Script, use Table -> Show Cheat Table Lua Script, paste the dofile(...) line below, and execute it:

dofile([[C:\path\to\cheatengine-mcp-bridge\MCP_Server\ce_mcp_bridge.lua]])

Look for: [MCP v12.0.0] MCP Server Listening on: CE_MCP_Bridge_v99

2. Configure MCP Client

Add to your MCP configuration (e.g., mcp_config.json):

{
  "servers": {
    "cheatengine": {
      "command": "python",
      "args": ["C:/path/to/MCP_Server/mcp_cheatengine.py"]
    }
  }
}

Restart the IDE to load the MCP server config.

For Codex, add a TOML server block to ~/.codex/config.toml:

[mcp_servers.cheatengine]
command = "python"
args = ['C:\path\to\cheatengine-mcp-bridge\MCP_Server\mcp_cheatengine.py']

Use single quotes for the Windows path so TOML treats backslashes literally.

TCP relay transport

The TCP relay lets mcp_cheatengine.py talk to Cheat Engine through a TCP socket instead of opening the Windows named pipe itself. Cheat Engine and the Lua bridge still run on Windows, while the MCP server can run anywhere that can reach the relay: another Windows process, a VM, a container, a Linux host, or a remote machine. This can also be used with WSL without changing the Lua bridge.

1. On Windows, load MCP_Server/ce_mcp_bridge.lua in Cheat Engine as usual.
2. On Windows, start the relay:

python C:\path\to\cheatengine-mcp-bridge\MCP_Server\ce_tcp_relay.py --host 127.0.0.1 --port 9876

3. In the environment where the MCP server will run, install the MCP dependency without pywin32, then run/configure the MCP server with TCP transport:

python3 -m pip install -r MCP_Server/requirements-tcp.txt

CE_MCP_TRANSPORT=tcp \
CE_MCP_HOST=127.0.0.1 \
CE_MCP_PORT=9876 \
python3 /path/to/cheatengine-mcp-bridge/MCP_Server/mcp_cheatengine.py

For MCP client configs that support environment variables:

{
  "servers": {
    "cheatengine": {
      "command": "python3",
      "args": ["/path/to/cheatengine-mcp-bridge/MCP_Server/mcp_cheatengine.py"],
      "env": {
        "CE_MCP_TRANSPORT": "tcp",
        "CE_MCP_HOST": "127.0.0.1",
        "CE_MCP_PORT": "9876"
      }
    }
  }
}

Set --host on the relay and CE_MCP_HOST on the MCP server to addresses that match your network setup. Keep the relay bound to trusted interfaces only, because anyone who can reach it can control the Cheat Engine bridge.

3. Verify Connection

Use the ping tool to verify connectivity:

{"success": true, "version": "12.0.0", "message": "CE MCP Bridge Active"}

4. Start Asking Questions

"What process is attached?"
"Read 16 bytes at the base address"
"Disassemble the entry point"

~180 MCP Tools Available

Memory

Analysis

Debugging

Process Lifecycle

Memory Allocation

Code Injection

Symbol Management

Assembly / Compilation

Window / GUI Automation

Input Automation

Cheat Table

Kernel Mode (DBK / DBVM)

And many more at AI_Context/MCP_Bridge_Command_Reference.md

Critical Configuration

BSOD Prevention

[!CAUTION] You MUST disable: Cheat Engine → Settings → Extra → "Query memory region routines"

Enabled: Causes CLOCK_WATCHDOG_TIMEOUT BSODs due to conflicts with DBVM/Anti-Cheat when scanning protected pages.

Troubleshooting

Cheat Engine says "too many local variables"

Load the bridge from disk with dofile(...) instead of pasting the full script into a cheat table script. The bridge also declares command handlers as global functions intentionally; this avoids Cheat Engine's Lua chunk limit of 200 local variables when the complete bridge is compiled at once.

MCP client cannot connect

Check these in order:

1. Cheat Engine is open and shows MCP Server Listening on: CE_MCP_Bridge_v99.
2. The MCP client was restarted after adding the server config.
3. The configured mcp_cheatengine.py path exists.
4. pip install -r MCP_Server/requirements.txt has installed both mcp and pywin32.
5. Run the MCP ping tool. A successful connection returns success: true and the bridge version. process_id: 0 is normal until Cheat Engine is attached to a target process.

Environment Variables

Example Workflows

Finding a value:

You: "Scan for gold: 15000"  →  AI finds 47 results
You: "Gold changed to 15100"  →  AI filters to 3 addresses
You: "What writes to the first one?"  →  AI sets hardware BP
You: "Disassemble that function"  →  Full AddGold logic revealed

Understanding a structure:

You: "What's at [[game.exe+0x1234]+0x10]?"
AI: "RTTI: CPlayerInventory"
AI: "0x00=vtable, 0x08=itemCount(int), 0x10=itemArray(ptr)..."

Project Structure

CLAUDE.md                               # Claude Code agent guidance (this repo)
README.md                               # User-facing documentation

MCP_Server/
├── mcp_cheatengine.py                  # Python MCP Server (FastMCP)
├── ce_mcp_bridge.lua                   # Cheat Engine Lua Bridge
└── test_mcp.py                         # Test Suite

AI_Context/
├── BATCH_WORKER_BRIEFING.md            # Parallel-worker task specifications (v12 overhaul)
├── MCP_Bridge_Command_Reference.md     # MCP Commands reference
├── CE_LUA_Documentation.md             # Full CheatEngine 7.6 official documentation
└── AI_Guide_MCP_Server_Implementation.md  # Full technical documentation for AI agent

Testing

Running the test:

python MCP_Server/test_mcp.py

Expected output:

✅ Memory Reading: 6/6 tests passed
✅ Process Info: 4/4 tests passed  
✅ Code Analysis: 8/8 tests passed
✅ Breakpoints: 4/4 tests passed
✅ DBVM Functions: 3/3 tests passed
✅ Utility Commands: 11/11 tests passed
⏭️ Skipped: 1 test (generate_signature)
────────────────────────────────────
Total: 36/37 PASSED (100% success)

The Bottom Line

You no longer need to be an expert. Just ask the right questions.

⚠️ EDUCATIONAL DISCLAIMER

This code is for educational and research purposes only. It's created to show the capabilities of the Model Context Protocol (MCP) and LLM-based debugging. I do not condone the use of these tools for malicious hacking, cheating in multiplayer games, or violating Terms of Service. This is a demonstration of software engineering automation.

内容由编辑整理，部分信息来自 GitHub 与项目官网，版权归原作者所有。发现错误？报告问题