Maintained by an individual developer (repository svnscha/mcp-windbg, MIT license); it is not an official Microsoft or MCP protocol project. It wraps cdb.exe / kd.exe behind ten MCP tools covering dump analysis, user-mode remote debugging, kernel debugging, bulk triage, and output filtering, with stdio and streamable-HTTP transports plus a Claude Code plugin and VS Code client config. It suits Windows developers doing crash and driver triage who already use an MCP client; it is not for non-Windows environments or users expecting automatic fixes. Before use, confirm the debugging tools are installed, understand that live and kernel sessions interrupt targets, and use a filter script to control sensitive data exposure.
A Model Context Protocol server that bridges AI models with WinDbg for Windows crash dump analysis, user-mode remote debugging, and kernel debugging. Under the hood it drives the Windows debuggers: CDB for user mode (dumps and -remote) and KD for kernel targets (-k). It is not a magical auto-fix, but a Python wrapper around cdb.exe / kd.exe that lets an LLM run real debugger commands and reason about the output. Every open_* call returns a session_id so several sessions (dumps, remote, kernel) can be open at once and addressed independently; per-call timeouts break into slow live commands with CTRL+BREAK and resynchronize the session. It runs locally over stdio or as a streamable-HTTP service you can drive from another machine, and a --filter-script can redact PII/secrets from tool arguments and output before they leave the machine.
{
"servers": {
"mcp_windbg": {
"type": "stdio",
"command": "python",
"args": ["-m", "mcp_windbg"],
"env": {
"_NT_SYMBOL_PATH": "SRV*C:\\Symbols*https://msdl.microsoft.com/download/symbols"
}
}
}
}