v1.0 · EN + PT · launch price

Make your AI write ABAP that actually activates.

Ready-made rules for Cursor, Claude Code, GitHub Copilot, Windsurf and any AGENTS.md agent, written by an ABAP developer and checked against SAP's own documentation.

Get the pack: 29 €Free lite version on GitHub
✗ Typical AI output
SELECT * FROM vbak INTO TABLE lt_vbak FOR ALL ENTRIES IN lt_keys " empty? reads ALL rows WHERE vbeln = lt_keys-vbeln. CALL FUNCTION 'Z_GET_ORDER_STATUS' " invented FM EXPORTING iv_vbeln = lv_vbeln. COMMIT WORK. " inside a RAP handler
✓ With ABAP for Agents
IF keys IS NOT INITIAL. SELECT vbeln, erdat, gbstk " S/4: status in VBAK FROM vbak FOR ALL ENTRIES IN @keys WHERE vbeln = @keys-vbeln INTO TABLE @DATA(orders). " 7.40 SP08+ ENDIF. " No COMMIT in RAP: the framework owns the LUW.

The problem: LLMs are confident and wrong in ABAP

Invented objects

Function modules, BAPIs and CDS views that don't exist or aren't released for ABAP Cloud.

Wrong release

INTO TABLE @DATA(...) on 7.40 SP05, += on 7.50, define view instead of view entities.

Silent data bugs

FOR ALL ENTRIES on an empty table reads the whole table. SELECT SINGLE without the full key.

RAP & cloud breakers

COMMIT WORK in behavior pools, sy-datum and WRITE in ABAP Cloud.

What's inside

10 focused rules

Core guardrails, syntax by release, ABAP SQL, ABAP Cloud & clean core A-D, RAP/CDS, Clean ABAP, obsolete statements, ABAP Unit, S/4HANA vs ECC, LLM self-check.

4 task skills

Create a managed RAP BO · refactor a legacy report to Clean ABAP · write ABAP Unit tests with test doubles · clean core code review.

Every tool, one source

AGENTS.md, CLAUDE.md + skills, Cursor .mdc, Copilot instructions + prompt files, Windsurf rules + workflows, plain prompts for ChatGPT / Joule chat.

Short on purpose

Research shows bloated context files hurt agents. Every always-on file stays under 200 lines; the rest loads by file type.

Sources included

Each rule maps to the Clean ABAP styleguide, the ABAP Keyword Documentation, SAP cheat sheets or the Cloudification Repository.

Generator included

Add your house rules (prefixes, namespaces, allowed BAPIs) and rebuild every format with one command.

Pricing

Launch price
29 €39 €
39 € after launch
  • All 10 rules + 4 skills
  • English and Portuguese
  • Cursor, Claude Code, Copilot, Windsurf, Codex
  • Generator + SOURCES.md
  • v1.x updates included
Get the pack

One-time purchase · free updates for v1.x · use in unlimited own and client projects. VAT may apply depending on your country.

Free lite (MIT)

0 €
  • 5 core rules + 1 review skill
  • EN + PT, all tool formats
  • Public GitHub repo
Download lite

Get notified about updates

New rules when SAP or the AI tools change: about one short email a month. No spam.

FAQ

Does this work with SAP Joule?

The rules target Cursor, Claude Code, GitHub Copilot, Windsurf and AGENTS.md agents. For Joule chat, use the included plain prompts. SAP does not document whether Joule reads rule files.

Does it work with ECC, or only S/4HANA and ABAP Cloud?

Both. The core rule makes the agent establish product, release and language version first, and the release table keeps 7.40/7.50 code free of newer syntax.

Can the AI still ignore the rules?

Yes. Rules are guidance, not enforcement. They are designed to cut repeat mistakes, but ATC and ABAP Unit stay your gate, and the rules tell the agent to run them via an MCP server when one is available.

Is it safe to use with the ADT MCP server?

The rules tell agents to use MCP read-only (look up objects, run checks) and never use write tools without your explicit consent.

Can I use it in client projects?

Yes, in unlimited own and client projects. You just can't resell or republish the pack itself.

Is this affiliated with SAP?

No. It's an independent product. SAP, ABAP and S/4HANA are trademarks of SAP SE.