discopop-mcp-server

<!-- This file is part of the DiscoPoP software (http://www.discopop.tu-darmstadt.de) Copyright (c) 2020, Technische Universitaet Darmstadt, Germany This software may be modified and distributed under the terms of the 3-Clause BSD License. See the LICENSE file in the package base directory for details. --> # DiscoPoP MCP Server A Model Context Protocol (MCP) server that exposes DiscoPoP functionality to MCP clients such as Claude Code and OpenCode, enabling an LLM agent to drive the full instrumentation pipeline and query static and dynamic data dependencies. ## Overview The DiscoPoP MCP Server bridges the gap between an LLM agent and DiscoPoP's profiling and analysis tools. It allows the agent to: - Instrument a project, run profiling, detect parallel patterns, and retrieve OpenMP patches - Query static and dynamic data dependencies for arbitrary code regions to support code understanding, refactoring, and correctness checks - Report the data a function and its callees were observed to read and write outside of it (side effects) - Retrieve profiling information from executed instrumented code - List and discover available profiling data and execution configurations ## Features - **Standalone CLI executable** - Run independently via command line - **Local deployment** - Uses stdio for direct integration with MCP clients (Claude Code, OpenCode, ...) - **Persistent daemon mode** - Optional long-running server that keeps analysis data in memory between calls - **Opt-in daemon** - Start `discopop_mcp_server --daemon` yourself; the server connects to it on the first tool call - **Transparent fallback** - Falls back to inline execution when no daemon is available - **Comprehensive logging** - View all incoming calls and outgoing responses - **Type-safe** - Full type hints throughout - **Extensible** - Easy to add new tools and capabilities ## Prerequisites - Python 3.11 or newer. - DiscoPoP installed in the **same environment** as the server. The tools call `…