Ghidra MCP Server

Instructions

Implementation Reference

• main.py:44-66 (handler)

  The core handler function for the 'setup_context' MCP tool. Decorated with @mcp.tool() for automatic registration and schema inference from type hints and docstring. Validates inputs, executes Ghidra headless analysis using helper functions, loads the resulting context, and updates global state.

  @mcp.tool()
  async def setup_context(ghidra_path: str = "/Users/tomi/Downloads/ghidra_11.3.1_PUBLIC", binary_path: str = "/Users/tomi/Documents/ghidra_mcp/crackme") -> str:
      """Run Ghidra headless decompilation to export binary context."""
      global ctx_ready, last_binary
  
      if not os.path.isdir(ghidra_path):
          return f"❌ Ghidra path '{ghidra_path}' is not valid."
      if not os.path.isfile(binary_path):
          return f"❌ Binary file '{binary_path}' does not exist."
  
      result = run_headless(ghidra_path, binary_path)
  
      if result.returncode != 0:
          return f"❌ Ghidra failed:\n{result.stderr}"
  
      if not is_context_ready():
          return f"❌ Export script ran but no context was saved."
  
      load_context()
      ctx_ready = True
      last_binary = os.path.basename(binary_path)
      return f"✅ Context loaded for '{last_binary}'."

• main.py:18-35 (helper)

  Key helper function called by setup_context to execute the Ghidra analyzeHeadless command with the export script, environment variable for JSON output path, and proper arguments for importing and analyzing the binary.

  def run_headless(ghidra_path: str, binary_path: str):
      analyze_headless = os.path.join(ghidra_path, "support", "analyzeHeadless")
      project_dir = os.getcwd()
      project_name = "ghidra_ctx"
      cmd = [
          analyze_headless,
          project_dir,
          project_name,
          "-import", binary_path,
          "-overwrite",
          "-scriptPath", os.path.dirname(EXPORT_SCRIPT_PATH),
          "-postScript", os.path.basename(EXPORT_SCRIPT_PATH),
          "-deleteProject"
      ]
      env = os.environ.copy()
      env["GHIDRA_CONTEXT_JSON"] = GHIDRA_CONTEXT_JSON
      return subprocess.run(cmd, capture_output=True, text=True, env=env)

• main.py:36-40 (helper)

  Helper function to load the exported Ghidra context JSON into the global ctx dictionary, used after successful headless run.

  def load_context():
      global ctx
      with open(GHIDRA_CONTEXT_JSON) as f:
          ctx = json.load(f)

• main.py:41-43 (helper)

  Helper function to check if the context JSON file exists, used to verify successful export.

  def is_context_ready():
      return os.path.exists(GHIDRA_CONTEXT_JSON)