CSPeach with an AI assistant and a connector

A connector gives an AI access to SAP. It does not make it an SAP developer.

Between access and a finished change sits a spec nobody questioned, a release the code has to suit, a house style, an order the objects must be built in, checks that have to pass, and a transport someone will read. That is what CSPeach is.

Where this starts. Connectors to SAP exist, and some of them are free. SAP ships an ABAP Development Tools MCP server inside ADT for Eclipse and for Visual Studio Code. Free open-source ADT connectors exist as well. What they give a coding assistant is access to ADT.

Facts about SAP's own tooling here are as of September 2026, from SAP's own release post.

The rest of the ticket

What CSPeach brings to it.

Each one ships today, on the systems listed at the foot of this page.

The questions no one asked
/abap-spec-gap reads the spec you already have and produces what it leaves out: the business questions a senior consultant would raise, the technical gaps, the edge cases and the test scenarios — before anyone writes a line.
35+ skills that carry written SAP rules
RAP and CDS, ABAP Cloud, clean core classified A to D, the S/4HANA upgrade scan, fix and verify, custom-code analysis across a whole estate, dumps and incidents, transports, Fiori apps. The rules live in the skill — naming, the order objects must be created and activated in, type codes, annotations, draft handling, access control — learned on live S/4HANA systems.
Code written for your release
On connect it reads the system's release and ABAP language version, and writes for that system. An ECC system is never handed syntax it cannot run.
Your standards file, on every turn
Put your team's ABAP conventions in a CSPEACH-STANDARDS.md file beside your code. Every turn carries them, and where they disagree with our defaults, yours win. Nobody re-explains the house style each morning.
The write road, fixed and always the same
A plan your developer approves, a snapshot, the write, a syntax check, activation, and a check that the object really went active — then its own transport. When a check fails, CSPeach fixes it and checks again; when it still cannot pass, it stops and asks. Broken code is never activated.
Write modes, set per system
advisory-only proposes and never writes; approval-gated asks before each write, and is the default; auto writes within the session for trusted non-production work. Tag a system as production and its mode is clamped to advisory-only for the whole session.
The audit export
/export audit writes the session record to a file your developer can hand to a reviewer: what was asked, what was written, what was checked, and where it went. Not a chat log.
A team project, with measured proof
A fix counts when a check proves it. The team project keeps the picture of the work in your own shared folder, and cspeach team refresh runs ATC to measure what has actually changed — so the progress report has a number behind it instead of a claim.
Seats, roles and keys in one place
The portal is where someone adds a developer, removes one, sets what each may do, and manages your own model key. Usage is shown in credits, per person and per skill.
Someone to call
When a live SAP system answers in a way nobody expected, there is a named team behind this, with SAP developers on it. Write to [email protected] and a person replies.

The road a change takes

The same road, every time.

This is the part a team has to trust before it lets an agent near a development system.

Every change takes the same road A write starts because you asked. For multi-object work the agent lays out a plan you approve, then it snapshots, writes, runs a syntax check, activates, and confirms the object really went active. If either check fails, CSPeach fixes the object and checks again, up to three tries, and takes the write again. When it still cannot pass, or the next step is one only a person can do, it stops and says what it needs; you do that step and it carries on from there. The whole road sits in its own transport and stays on the record. THE PATH OF A WRITE Every change takes the same road. ITS OWN TRANSPORT · R9 ON THE RECORD · /export audit PASSES FAILS FAILS TRIES AGAIN STILL FAILING, OR IT NEEDS YOU YOU DO THE STEP · IT CARRIES ON You ask R8 A plan you approve MULTI-OBJECT WORK R7 Snapshot Write Syntax check R10 · EVERY WRITE Activate Activation check REPORTS IF IT DIDN'T It stops and asks you SAYS WHAT IT NEEDS CSPeach fixes it and checks again UP TO THREE TRIES Broken code is never activated. The snapshot from before the write is there if you want to go back.
The road one change takes, from the ask to a transport of its own with the record of what happened.

Stage by stage

What CSPeach does, and what is left to you.

One ticket, from the spec on someone's desk to a change a reviewer can read.

Stage of a ticket CSPeach An assistant plus a connector
Before any code /abap-spec-gap produces the question list, the edge cases and the test scenarios from the spec you already have. You work out what the spec leaves open and take the questions to the business yourself.
The plan /abap-design lists the objects and the order they must be built in; /abap-plan holds the phases across sessions. You break the work into objects and put them in build order in your head or in a ticket.
Writing the code It writes for the release it read from your system, and follows your CSPEACH-STANDARDS.md on every turn. The assistant writes from general ABAP knowledge, and you correct it for your release and your house style.
Before the write A snapshot is taken first, every time, and the session can be rewound to it. You remember to keep a copy of what is there now.
The checks A syntax check on every write, activation, then a check that the object really went active — in that order, without being asked. ATC on the change with /abap-atc-fix, and across your custom code in the upgrade and clean-core skills. You run the syntax check and ATC and read the findings.
A failed check It fixes the object and checks again. When it still cannot pass, it stops and says what it needs. You paste the finding back into the chat and try again.
The transport The change goes in its own transport, and /export audit writes the record for whoever reviews it. You choose the transport and keep it clean.
More than one developer A team project in your own shared folder, where a fix counts when ATC proves it. Everyone has their own chat history, and progress is what people say it is.

Facts about SAP's own tooling on this page are as of September 2026, from SAP's own page linked above.

Questions

Asked plainly, answered plainly.

Why pay when the connector is free?

Because you are paying for the work either side of the connection: the questions before the code, code written for your release and your standards, the fixed road every write takes with the checks run and the failures fixed, a transport of its own, a record a reviewer can read, and one measured picture of the work when more than one developer is on it.

Does CSPeach need SAP's MCP server?

No. CSPeach talks to SAP over ADT directly from the laptop, the same protocol Eclipse uses. Nothing extra is installed on the SAP system, and nothing extra has to be switched on in an editor.

How CSPeach reaches SAP

Which systems does it run on?

ECC EHP7 and higher, on-premise S/4HANA, and S/4HANA Cloud Private Edition. S/4HANA Cloud Public Edition and SAP BTP ABAP Environment: not tried yet.

Does it touch our production system?

No. CSPeach is a development agent: it works in your development system, and it is not connected to production. A change reaches production the way all your changes do — through your own transports and your release process. If someone does connect it to a system tagged production, its write mode is clamped to advisory-only: it proposes and never writes.

How the write gates work

Whose key, whose prompts?

Ours by default, through our proxy to the model provider. With your own model key the AI call goes straight from the laptop to the model provider, and we hold that key encrypted. Sign-in and the signed skills reach us either way, and ADT traffic never passes through us.

Code, credentials and prompts

Can we try it before we decide?

Yes, and that is what we would rather you did than read another page. One week, one real ticket, on your own system, with your developer driving. Write to [email protected] and we will set it up.

We also have SAP Joule for Developers.

Then the other comparison page is the one to read. It sets out the work CSPeach does on the system you have today, and what each one needs before you start.

CSPeach for a team that already runs Joule

One week, one real ticket, on your own system.

Your developer drives. We watch, and we answer. At the end you have a change you can read.