oisee/vibing-steampunk

vs-punk: ADT to MCP bridge - Vibe code in ABAP / AMDP

What it solves

vibing-steampunk (vsp) provides a bridge between AI agents (via the Model Context Protocol - MCP) and SAP ABAP systems. It allows AI assistants like Claude to read and write code, debug, deploy, and run tests on SAP systems using natural language or automation scripts, eliminating the need for manual navigation of SAP's complex internal tools.

How it works

The tool acts as an ADT (ABAP Development Tools) ↔ MCP Bridge. It leverages SAP's own ADT APIs over either a pinned classic-RFC conversation or a stateful HTTPS session. Crucially, it requires nothing to be installed on the SAP server itself.

Key technical mechanisms include:

  • Session Management: It maintains a stable ABAP session to handle complex operations like debugging, which would otherwise fail across independent HTTP calls.
  • Protocol Translation: It translates MCP tool calls into ADT resource requests (e.g., /sap/bc/adt/debugger/stack).
  • Data Decoding: It implements SAP-specific decompression (LZH and LZC) in Go to read cluster tables (like BALDAT, INDX, STXL) and the MIME repository without needing ABAP code on the server.
  • DAP Integration: It includes a Debug Adapter Protocol (DAP) server, allowing standard editors like VS Code and Neovim to set breakpoints and step through ABAP code.

Who it’s for

  • ABAP Developers: Who want to integrate AI agents into their development workflow or use modern editors (VS Code, Neovim) for debugging.
  • SAP Consultants: Who need to perform post-mortem analysis of system dumps and application logs without manual digging.
  • SAP System Administrators: Who encounter issues with jobs and spool requests that need to be exported or analyzed.

Highlights

  • Zero-Server Installation: Operates entirely via ADT APIs, requiring no Z-code or SDKs on the SAP server.
  • Full Debugger Access: Supports breakpoints, attaching, stepping, and reading/writing variables in real-time.
  • AMDP Debugging: Bridges ABAP and HANA debuggers to allow debugging of SQLScript methods.
  • Post-Mortem Analysis: Automatically correlates system dumps (ST22) with application logs to suggest root causes.
  • Advanced Data Access: Decodes cluster tables and the MIME repository byte-exactly using Go-native logic.
  • Response Caching: Uses a Go-native SQLite cache to reduce response times from seconds to milliseconds.

Related

  • Project
    dbt-labs/dbt-mcp

    dbt‑MCP Server is an open‑source service that exposes a large catalog of dbt‑related tools (SQL execution, semantic‑layer queries, model discovery, dbt CLI commands, job administration, code generation, and documentation lookup) via the Model Context Protocol. It lets LLM‑powered agents understand and act on a dbt project—generating SQL from natural language, running dbt jobs, fetching lineage, and even creating new model files—making it a concrete bridge between data‑engineering metadata and AI assistants.

  • Project
    microsoft/azure-devops-mcp

    An MCP server that gives AI agents direct access to Azure DevOps data and tools, enabling them to manage work items, repositories, and pipelines.

  • Project
    HelpCode-ai/anythingmcp

    AnythingMCP is an open‑source, self‑hosted gateway that converts REST, SOAP, GraphQL, OData, or SQL services into MCP “tools” usable by Claude, ChatGPT, Copilot and other LLM clients. It ships with 352 pre‑built connectors (SAP, Etsy, Amazon Seller, etc.), a visual editor for field‑level privacy, RBAC/SSO, audit logging, and can be run via a one‑line Docker‑compose command or deployed as a managed cloud service.

  • Project
    miscusi-peek/cheatengine-mcp-bridge

    A bridge that connects AI agents to Cheat Engine via the Model Context Protocol, enabling natural language interaction with program memory for reverse engineering and modding.

  • Project
    svnscha/mcp-windbg

    An MCP server that bridges AI models with WinDbg, enabling natural language crash dump analysis, remote user-mode debugging, and kernel debugging.