ude-mcp

MCP (Model Context Protocol) server for the PLS UDE® (Universal Debug Engine) debugger. It lets AI coding agents drive UDE over the official UDE Object Model automation API: load a workspace, connect a target, flash an ELF, run/halt/step, set breakpoints, and read back variables, registers, memory, and call stacks.

Status

✅ Working prototype — end-to-end verified against the built-in TSIM simulator (session start → connect → register/memory reads → run control → clean shutdown). Real-hardware (AURIX TriCore via DAP/JTAG) testing is next.

How it works

UDE Starterkit cannot be automated from outside its process (the COM launcher is not shipped/registered in the free toolchain), so ude-mcp injects a JScript bridge agent into UDE as a startup macro (UDE.exe -s<agent.js>). The agent binds to the live UDE object model and talks to the MCP server over localhost HTTP long-polling:

AI client ⇄ (MCP/stdio) ⇄ ude-mcp (Python) ⇄ (localhost HTTP) ⇄ bridge.js (in UDE) ⇄ UDE object model ⇄ TSIM / target

Everything uses officially supported UDE mechanisms (startup macros + the COM object model); no patching or unsupported hacks. See docs/api-selection.md for the full API assessment.

Related MCP server: gdb-mcp

Target platform

• Windows only (UDE is a Windows application; the automation API is COM-based)

• UDE Starterkit / UDE installed with the HighTec TriCore toolchain

• First milestone: drive the built-in TSIM simulator (no hardware needed), then real AURIX TriCore targets over DAP/JTAG

Planned MCP tools

Development

uv sync
uv run ude-mcp   # stdio MCP server

Smoke test (launches a real UDE instance against the TSIM simulator):

uv run python scratch/e2e.py

MCP client configuration

{
  "mcpServers": {
    "ude": {
      "command": "uv",
      "args": ["--directory", "C:\\sw\\ude-mcp", "run", "ude-mcp"],
      "env": { "UDE_MCP_UDE_EXE": "C:\\Program Files\\pls\\UDE Starterkit 2021\\UDEVisualPlatform.exe" }
    }
  }
}

License

MIT