Skip to main content
pwno-mcp includes automation tools for build steps and helper services, but they are intentionally separate from debugger control.

run_command

Use for compiles, build scripts, and short-lived helper commands.

spawn_process

Use for long-running helper processes such as local HTTP servers or listeners.

execute_python_code

Use for quick Python probes instead of python -c.

fetch_repo

Use for cloning helper repositories into /workspace.

​
Example flow

1

Build a target or helper binary

{"tool":"run_command","arguments":{"command":"gcc -g vuln.c -o chal","cwd":"/workspace","timeout":30.0}}
2

Start a background helper service

{"tool":"spawn_process","arguments":{"command":"python3 -m http.server 8000","cwd":"/workspace"}}
3

Inspect or stop the helper

{"tool":"get_process","arguments":{"pid":12345}}
{"tool":"kill_process","arguments":{"pid":12345,"signal":15}}

​
Guardrails

  • do not use run_command or spawn_process to start the target ELF you want to debug
  • do not use python -c inside run_command; use execute_python_code instead
  • use pwncli when you need interactive exploit-driver I/O instead of a generic subprocess
The subprocess backend actively rejects some unsafe usage patterns and returns a ToolUsageError-style payload with a recommended tool when you pick the wrong interface.