Person
Marian Zeis
Capabilities
Provides
Agent access to your codeABAPCAPUI5BTPAgent access to your codeLetting a coding agent read the project or the system you actually have, reach its model, its objects and its checks, and act on it. In file-based environments that works out of the box; complex systems need dedicated access capabilities. Whatever the environment, the agent also needs project- and framework-specific context, so it understands the thing in front of it rather than guessing from what it remembers of its training data. The transport is incidental: an MCP server, a language server or a CLI answer the same question, and what differs between them is how much of the real thing the agent can see. Distinct from [MCP servers](capability:mcp-servers), which is a protocol adapter on your own running service, and from [agent guidance](capability:agent-guidance), which is knowledge handed over as files rather than read from the thing itself.Agent evalsABAPCAPBTPAgent evalsKnowing whether an AI feature still behaves after a model, prompt or schema change. Regression testing for non-deterministic output, against fixtures that reflect real business data.Agent guidanceABAPCAPUI5BTPAgent guidancePackaged knowledge you hand a coding agent so it gets a framework right unprompted — skills, rule files, instruction packs. Distinct from agent tooling, which is about building agentic features into your application, and from [agent access to your code](capability:agent-code-access), where the agent reads the real project instead of being handed a file about it.Browsing table dataABAPBrowsing table dataLooking at what a system holds while developing or supporting it: the rows behind a key, the reference data a field points at, the tables that carry data for one object. Editing configuration records through maintenance screens is a different job, even where the same transaction does both.DocumentationABAPCAPUI5BTPDocumentationFinding the answer in what has already been written, and writing it where it has not — first-party documentation made searchable by something other than a person scrolling it, including by an agent, and a document of custom code produced from the code itself.Editor supportABAPCAPBTPUI5Editor supportWorking in source code with the editor helping rather than treating it as text — completion, go-to-definition, inline errors, refactoring and preview.Tool governanceABAPCAPBTPTool governanceControlling what an agent may call, with which authorisations, and leaving an audit trail of what it did. The authorisation question that MCP exposure raises but does not answer.
Demonstrates
Agent access to your codeABAPCAPUI5BTPAgent access to your codeLetting a coding agent read the project or the system you actually have, reach its model, its objects and its checks, and act on it. In file-based environments that works out of the box; complex systems need dedicated access capabilities. Whatever the environment, the agent also needs project- and framework-specific context, so it understands the thing in front of it rather than guessing from what it remembers of its training data. The transport is incidental: an MCP server, a language server or a CLI answer the same question, and what differs between them is how much of the real thing the agent can see. Distinct from [MCP servers](capability:mcp-servers), which is a protocol adapter on your own running service, and from [agent guidance](capability:agent-guidance), which is knowledge handed over as files rather than read from the thing itself.DocumentationABAPCAPUI5BTPDocumentationFinding the answer in what has already been written, and writing it where it has not — first-party documentation made searchable by something other than a person scrolling it, including by an agent, and a document of custom code produced from the code itself.
Contributes to
Agent access to your systemsABAPBTPCAPAgent access to your systemsGiving an AI agent access to the business data, configuration or operations of systems you run through an MCP server. [Serving your application over MCP](capability:mcp-servers) exposes your own service contract; system access reaches the systems behind it, and one implementation may do both. Development assistance belongs to [agent access to your code](capability:agent-code-access).Analytic modellingABAPBTPAnalytic modellingDefining what an analytical client actually reads — the multi-dimensional layer between stored rows and a chart: measures, dimensions, hierarchies, filters and parameters, and the associations somebody navigates them by. Distinct from [analytical queries](capability:analytical-queries), which is a service answering the request a chart sends, and from [loading a warehouse](capability:warehouse-loading), which fills the tables underneath it.Loading a warehouseABAPBTPLoading a warehouseGetting data into an analytical store and keeping it current — connections to source systems, replication and the other integration tasks, import jobs and their lifecycle, and the monitoring that tells you a load failed before a report does. Distinct from [replication](capability:replication), which is an application reading data that lives elsewhere as if it were local rather than a warehouse being filled.LocalizationABAPCAPBTPUI5LocalizationServing labels, messages and translated content per user locale, including translatable model elements and per-tenant text overrides.
Entries
Group by20 of 20 entries
EntrySignals
Maintains
@arc-mcp/adt-ls@arc-mcp/adt-lsA TypeScript SDK over adt-ls, the headless language server SAP ships inside ABAP Development Tools for VS Code, so a program drives it without the discovery, JVM and handshake work.independent, abap
What adt-ls cannot do headless the maintainer puts out of scope, so a consumer that needs an ADT call the language server does not expose is back to speaking ADT itself. The README states the current runtime compatibility as verified on macOS arm64, against adt-ls 1.1.2 as the minimum and `1.1.2.202608131517` as the version tested. An Eclipse installation does not supply that runtime: SAP versions the two separately, so the VS Code extension has to be installed even where nobody opens VS Code.
TypeScript SDK over SAP's headless adt-ls language server
arc-1arc-1An MCP server in front of the ADT REST API that runs locally and on Cloud Foundry with twelve tools that read, search, write, activate, lint and transport ABAP objects.independent, abap
MCP server over the ADT REST API for ABAP objects
arc-1-lsparc-1-lspAn edition of ARC-1 that hands every ADT interaction to SAP's own embedded adt-ls language server instead of a hand-rolled HTTP client.independent, abap
Single-tenant today, one technical user, with per-user principal propagation on the roadmap — so a shared deployment cannot yet act as the developer who asked. The sibling edition [ARC-1](artifact:arc-mcp-arc-1) speaks ADT directly if you would rather not depend on the language server.
Edition of ARC-1 that runs ADT calls through adt-ls
arc1-adt-abap-mcp-extarc1-adt-abap-mcp-extOne JAR in Eclipse's dropins that turns on SAP's own ADT MCP server and adds read-only repository tools beside it — so an AI client reads your system without leaving Eclipse.independent, abap
It lives inside Eclipse and inherits everything that implies — the session is a developer's, not a pipeline's. And it is version-bound: ADT 3.60 or newer for 0.5.1, with 0.3.x the path for ADT 3.58 and 3.59.
Eclipse add-on enabling the ADT MCP server with read-only tools
mcp-sap-docsmcp-sap-docsAn MCP server over seventeen cloned SAP documentation repositories — ABAP keyword docs, capire, OpenUI5, Cloud SDK, BTP — searched with BM25 and prebuilt embeddings fused by reciprocal rank fusion.independent, cap
It reads documentation, never a system: nothing in it opens a package.json, a CDS model or an ABAP client, so a question about what your own project does has to be answered elsewhere. The offline half is only as fresh as the submodules the last build pinned, and the index answers where something is written down rather than which of two packages fits a task.
MCP server searching seventeen SAP documentation repositories
SAP API Policy Evidence SkillSAP API Policy Evidence SkillAssesses whether a described SAP API usage aligns with the SAP API Policy, gathering live evidence from SAP's own sources and returning a sourced assessment with a stated confidence — and routing the legal questions away from itself.independent, btp
It says so itself, in bold, and the atlas repeats it: not legal advice and not a compliance decision. Only SAP, your contract, or SAP support can answer bindingly, and the skill routes legal, commercial and roadmap questions there rather than guessing at them. Treat the output as evidence gathered, not as a verdict.
Skill assessing API usage against the SAP API Policy
sap-note-search-mcpsap-note-search-mcpAn MCP server over SAP Notes and Knowledge Base articles: search finds a note, fetch returns it whole, and a scripted browser does the S-user login the content sits behind.independent, btp
It needs an S-user with SAP Support Portal access, so there is nothing to search without a support contract behind the login. The maintainer states it uses private SAP APIs behind authentication, asks the reader to check whether that use breaches SAP's terms of service, and disclaims both liability and any guarantee that it keeps working. MFA is handled only by typing the code into a visible browser, so a headless deployment cannot complete a login that demands one.
MCP server searching and fetching SAP Notes and KBAs
WerkbankWerkbankA local desktop workspace for AI conversations, attached files and generated HTML Spaces, run by a goose agent and offering a catalog of SAP MCP connectors and ABAP, CAP and UI5 skills.independent, umbrella
Local desktop workspace for AI conversations with SAP connectors
ztoadztoadAn Open SQL editor for ABAP that runs a query written in the editor pane on the server and shows the rows in an ALV window, with a pane of saved queries to reuse and share.independent, abap
Open SQL editor for ABAP with ALV output
Maintains: samples, sketches and references
abap-docsabap-docsSAP's ABAP Keyword Documentation scraped from help.sap.com and committed as Markdown in two libraries, Standard ABAP and ABAP Cloud, with keyword and category indexes for the MCP servers and LLMs it is generated for.independent, abap
The Markdown is a snapshot of whatever help.sap.com served when the pipeline last ran, and the Standard library is scraped from the `latest` index, so which release a page describes follows from the scrape date rather than from anything the corpus states. `package.json` states ISC and the repository carries no LICENSE file, so what the copied SAP documentation may be used for is not stated in it.
ABAP Keyword Documentation as Markdown with indexes
LLM-Benchmark-ABAP-Code-GenerationLLM-Benchmark-ABAP-Code-GenerationA benchmark that has a model write ABAP for 180 tasks ten times each, then feeds the SAP system's own syntax check, activation and ABAP Unit results back for up to five correction rounds.independent, abap
Benchmark of LLM ABAP code generation against a real SAP system
local-llm-client-for-sap-consultants-librechatlocal-llm-client-for-sap-consultants-librechatA fork of LibreChat that ships four SAP MCP servers pre-configured: documentation, SAP Notes, S/4HANA OData and ABAP ADT, so the chat client a consultant runs on their own machine can reach a customer system.independent, abap
LibreChat fork with four SAP MCP servers preconfigured
Contributes to
ARC-1 SkillsARC-1 SkillsTwenty-two SKILL.md files shipped in arc-1's own skills/ directory for RAP scaffolding, unit test generation, clean-core audits and migration paths, loaded on demand by description and namespaced /arc-1:<skill> when installed through the Claude Code plugin.independent, abap
Every skill that touches ABAP needs [arc-1](artifact:arc-mcp-arc-1) itself connected to a real system; nothing else is universal. `modernize-ui5-app` and `convert-ui5-to-fiori-elements` specifically need [the UI5 MCP server](artifact:ui5-mcp-server) for TypeScript conversion guidance and the linter, `convert-ui5-to-fiori-elements` also needs [the Fiori MCP server](artifact:sap-ux-fiori-mcp-server) for its LROP scaffold, and `modernize-ui5-app` needs a browser MCP (Claude in Chrome or Claude's own Preview) for its final render-verification step, since an HTTP 200 alone does not catch a blank page. A skill file alone, without the servers a given skill needs, has nothing to act on.
Twenty-two agent skills shipped with arc-1
arc-1-extension-samplearc-1-extension-sampleA sample arc-1 extension in TypeScript, with no ABAP, whose ten Custom_* tools each show one way the plugin API reaches SAP: ADT and OData reads, an OData write, class and report execution, a custom ICF service and classic RFC.independent, abap
No license Neither the package manifest nor the repository tree declares a licence.
Functional boundary It consumes the `@experimental` public API, which the README says may break in any release. The RFC tool dials as one shared technical user with its own credentials, so SAP's log records that user and not the person behind the agent.
Sample arc-1 extension in TypeScript with ten custom tools
arc-mcp-hubarc-mcp-hubA hub on SAP BTP that puts several arc-1 instances, one per SAP system, behind one login and a route each, exchanging the caller's token through a destination so each system still sees the real user.independent, btp
Archived The declared repository is archived.
Functional boundary With one SAP system the README says to point the client at that arc-1 directly. Version 1 exchanges tokens within a single subaccount, so backends spread across subaccounts are out; it keeps its session map in memory, so more than one instance needs sticky sessions or a shared store. It routes deterministically and runs no model, which is a different thing from an assistant that reasons across systems. The `/all` endpoint drops the per-connection isolation, and the maintainer's answer to that is a PROD backend set read-only at the SAP end rather than a tool surface that hides the writes.
Hub on SAP BTP putting several arc-1 instances behind one login
arc1-abap-bridgearc1-abap-bridgeA VS Code extension between arc-1 and ABAP Development Tools for VS Code, so GitHub Copilot can open an ABAP object by name, diff the inactive version and ask the system whether an edit would be accepted.independent, abap
The diff holds only while the object is inactive, since after activation the previous version is gone and a watched journal directory is what the maintainer points to instead. It adds no write path of its own: the IDE writes through SAP's extension, and the README states arc-1 can stay on `SAP_ALLOW_WRITES=false`. DDIC metadata files are skipped, because SAP's own stub for them says they are not supported in ABAP development tools for VS Code.
VS Code bridge between arc-1 and ADT for VS Code
btp-sap-odata-to-mcp-serverbtp-sap-odata-to-mcp-serverExposes every OData service on an S/4HANA or ECC system as MCP tools, in three discovery levels so a large catalogue does not have to be loaded at once.independent, btp
It puts create, update and delete on business entities behind natural language. Whatever the destination's principal may change, an agent may change — and "create a new customer" is a sentence, not a form somebody signed off.
Exposes S/4HANA or ECC OData services as MCP tools
bw-modeling-mcpbw-modeling-mcpReads, creates and modifies BW/4HANA modelling objects through the same internal APIs Eclipse BWMT and the BW/4HANA Cockpit use — and can run centrally on BTP with each caller reaching BW as themselves.independent, btp
Its README says it plainly and so does this entry: this is not a simulation, every tool call reaches a live BW system, and write operations produce real changes. BW modelling objects are the definition of a warehouse — the blast radius of a wrong one is every report built on it.
MCP server for BW/4HANA modelling objects
LISALISAAn MCP server that lets an AI client read and write the translations of ABAP repository objects through three tools, backed by an ABAP HTTP service on SAP's XCO i18n APIs.independent, abap
LISA does the translation step only: it does not discover objects or manage transports, and its README pairs it with [an ADT MCP server](artifact:sap-adt-mcp-server) for those. Every write needs a transport request, and which object types it can write differs between public cloud or [the SAP BTP ABAP environment](artifact:btp-abap-environment) and on-premise or private cloud, and by release. Running it locally is meant for development and testing, not production.
MCP server for reading and writing ABAP object translations
vibing-steampunkvibing-steampunkAn ADT-to-MCP bridge in Go that reaches past what ADT exposes — cluster tables decoded, spool and job logs read as tables, the IMG searched, selection texts written back.independent, abap
Reach is the reason to use it and the reason to be careful: reading cluster tables and spool output goes well past what a development tool is normally pointed at. Decide what an agent may reach before granting it, not after.
ADT-to-MCP bridge in Go reaching past what ADT exposes
Writing and topics
14 of 14 pieces
Piece, what it offersStill applies
AI-assisted SAP development with MCPAI-assisted SAP development with MCPA Devtoberfest 2026 session (UI-41) that traces AI-assisted SAP development from pasting code into a chat window to coding agents with built-in connectors, and sorts the MCP servers around SAP development into SAP's own, the community's and the hosted ones, where the agent cannot change what the server allows. Four demos follow: GitHub Copilot alone explaining an ABAP report from a Git repository, @ui5/mcp-server reading an OpenUI5 sample app and converting it to TypeScript, mcp-sap-docs grounding Terraform files for a BTP subaccount the presenter had not written before, and arc-1 with Copilot in Eclipse tracing why a title longer than 40 characters is rejected and proposing the DDIC-allowed length instead. It ends on a skill that migrates an SAP Gateway project on an ECC system to RAP on S/4HANA step by step, and on a rule Marian Zeis gives for servers written for internal use: keep them read-only, because what the agent is short of is context.published 2026-10-16
ARC-1 1.0: One Deployment, Multiple SAP SystemsARC-1 1.0: One Deployment, Multiple SAP SystemsThe release post for ARC-1 1.0, whose largest addition is an experimental multi-system mode: one BTP Cloud Foundry deployment attaches the SAP systems or clients an administrator selects through destinations, each with its own MCP address and an optional combined one, beside 22 included skills, an extension framework for company-specific tools, the BTP login code split out as @arc-mcp/xsuaa-auth, and a move to the ARC MCP organisation. It names the contributions behind the release: Wouter Lemaire for the first Cloud Foundry deployment and the first multi-backend implementation, Sami Bouguerra for NetWeaver 7.50 fixtures and activations that looked successful but were not, Clément Ringot for BTP login and Cloud Connector fixes, and Geert-Jan Klaps for S/4HANA Public Cloud support. It places SAP's own ADT MCP Server as the choice for a developer inside VS Code or Eclipse and ARC-1 as a shared service on BTP, and it stops short of a licence answer: the ADT web endpoints are not listed as published APIs in the SAP Business Accelerator Hub, so readers are told to ask SAP whether their agreement allows third-party tools to use them before a broader or production rollout.published 2026-08-03
A Clearer View of Your SAP Integrations Under the New API PolicyA Clearer View of Your SAP Integrations Under the New API PolicyThe SAP API Policy Evidence Skill assesses one SAP interface or usage pattern at a time against SAP API Policy v.4.2026a, reading evidence from five MCP servers, among them servers for the Business Accelerator Hub and SAP Notes, mcp-sap-docs and a read-only ARC-1, and returns one of four labels with a confidence level, an evidence table, residual risk and questions for SAP. It gives no compliant or not compliant answer on purpose, because SAP's FAQ (Q49) declines to publish a binding decision matrix, and it sends legal, commercial and roadmap questions to SAP and lowers its confidence when a source is not connected. Four sample runs show the method, among them ODP-RFC extraction labelled likely not aligned under SAP Note 3255746 and the same API_SALES_ORDER_SRV assessed differently for an autonomous agent behind a custom gateway and for a Boomi middleware flow; the SAP Notes and Business Accelerator Hub servers read SAP sites through a browser login, which the post says to use at your own risk.published 2026-06-02
Before You Upgrade S/4HANA: Static UI5 Checks with MCPBefore You Upgrade S/4HANA: Static UI5 Checks with MCPA UI5 version upgrade checked against one app by pairing two MCP servers: the ui5_version_diff tool added to mcp-sap-docs, which reads a bundle from ui5-lib-diff downloaded once at setup and filters it by version range, change type, library or query text, and @ui5/mcp-server, whose get_project_info, get_version_info, run_ui5_linter and run_manifest_validation say what the app uses and what fails. Run with a ui5-versions-upgrade.md skill on the legacy app from From SEGW and Legacy UI5 to RAP with ARC-1, from SAPUI5 1.84.59 to 1.147.2, the diff range held 985 deprecations and 961 What's New entries, and the linting and manifest validation narrowed that to findings in this app, among them jQuery.sap.require, synchronous bootstrap, the sap_belize theme, sap.m.MessagePage and an old manifest _version. The checks read the source in a local checkout and never the deployed app, so the report is a first pass: a changelog entry is not proof that a workaround can be deleted, and the app still has to be run and tested.published 2026-05-20
ABAP Development Automation with ARC-1 and GitHub WorkflowsABAP Development Automation with ARC-1 and GitHub WorkflowsA sample repository puts an ABAP package into GitHub with abapGit and connects GitHub Actions to ARC-1 on BTP through an API key with no write tools, so abaplint on the diff, ABAP Unit and ATC through SAPDiagnose, label-triggered Copilot and Claude reviews, and ST22 dump triage all run as checks on a pull request. The review case is PR #14, where a SAPRead of the activated method let Claude state that a direct SELECT from MARA was new in the PR and not yet active in SAP, which a diff-only review cannot know. The test gate reads activated objects, so a team that pushes before activation needs a sandbox or a pull-and-activate step, and pulling a merged change back into SAP is named as the next step, not built.published 2026-05-12
From SEGW and Legacy UI5 to RAP with ARC-1From SEGW and Legacy UI5 to RAP with ARC-1Three ARC-1 skills, run in Cursor with Composer-2 against a deliberately rough demo on an S/4HANA 2023 trial system, turn the SEGW service ZDEMO_MIG_PROJECTS_SRV into a managed, draft-enabled RAP business object published as OData V4, replace its ApproveProject function import and its direct UPDATE ... COMMIT WORK with the action approve_project written in EML in local mode, and rebuild the freestyle UI5 JavaScript app once in UI5 TypeScript and once as a Fiori elements list report and object page driven by metadata extensions. What the published transcripts add is how the agent got there: it found the generated MPC class names and the custom logic in DPC_EXT only by reading the system, the skills made it print the extracted model and stop for confirmation before writing, and @ui5/mcp-server, mcp-sap-docs and @sap-ux/fiori-mcp-server each had their own job beside it. The demo is small on purpose, and the post says a productive SEGW service that calls BAPIs, carries custom authorisation logic or supports deep inserts is not solved with one prompt.published 2026-05-11
From SEGW and Legacy UI5 to RAP with ARC-1
@sap-ux/fiori-mcp-server has been through 2 major lines since this was published: 0.7 (2026-05), 1 (2026-05). Affects Runnable code.
ARC-1 with Joule Studio: Bringing Real ABAP System Context into JouleARC-1 with Joule Studio: Bringing Real ABAP System Context into JouleBrings the same BTP-deployed ARC-1 endpoint into Joule through a Joule Studio, classic edition agent configured with Anthropic Claude Sonnet 4 and two MCP servers, ARC-1 and mcp-sap-docs, for work that starts in SAP where the previous part started in Microsoft 365. Two read-only showcases carry the ABAP Cloud readiness edge: a clean core check of one class that reports it is not released, uses USR02 and BUT000 directly and has no ATC findings, and a modernisation backlog for package ZSD_LEGACY which, as that package did not exist, grouped 28 ZSD* objects into risk buckets instead. The route through SAP Integration Suite API Management or a future MCP gateway is presented as a target architecture and not yet a productive setup, because Joule Studio's documentation requires streamable HTTP destinations and does not support MCP servers that need interactive OAuth user authorisation.published 2026-05-06
ARC-1 with Joule Studio: Bringing Real ABAP System Context into Joule
arc-1 has been through 2 major lines since this was published: 0.9 (2026-05), 1 (2026-07). Affects Runnable code.
ARC-1 with Copilot Studio: SAP System Context Beyond DevelopersARC-1 with Copilot Studio: SAP System Context Beyond DevelopersTakes the BTP-deployed ARC-1 endpoint into Microsoft Copilot Studio, added as an existing MCP server with OAuth 2.0 against XSUAA and published to Teams or Microsoft 365 Copilot, so SAP system context reaches project roles that do not work in an IDE. The test agent combines SharePoint specifications, Jira tickets and mcp-sap-docs with ARC-1 in six scenarios: a patch for a Jira bug proposed before it is applied, the impact of widening a domain from CURR 13,2 to CURR 23,2, a two-layer CDS view built from a specification, an ATC and abaplint quality report, a short dump diagnosis, and a clean core check that proposes released successors for direct selects on USR02 and BUT000. The access path and security model are those of the BTP part unchanged; the advice is to start this audience read-only, although the showcase itself enabled the read and write tools.published 2026-05-04
ARC-1 with Copilot Studio: SAP System Context Beyond Developers
arc-1 has been through 3 major lines since this was published: 0.8 (2026-05), 0.9 (2026-05), 1 (2026-07). Affects A recipe, Runnable code.
ARC-1 on SAP BTP: Secure ABAP Agentic Development Beyond the LaptopARC-1 on SAP BTP: Secure ABAP Agentic Development Beyond the LaptopThe BTP deployment of ARC-1: one Cloud Foundry app per SAP system, reached by MCP clients through XSUAA and reaching the backend through the SAP Destination service, the SAP Connectivity service and Cloud Connector, with a BasicAuth destination for startup probing and a Principal Propagation destination for per-user requests, and SAP_PP_STRICT=true so a propagation problem fails instead of falling back to the shared user. Tool governance is the three gates it spells out, the server ceiling set by environment variables, the user's permission from XSUAA role collections, OIDC scopes or API-key profiles, and the SAP backend authorisation, all of which must allow an action, and the staged rollout from read-only to writes in $TMP, then selected packages, then transport writes. It recommends the MTA deployment over Docker on Cloud Foundry or a buildpack push, treats BTP ABAP Environment as a separate service-key scenario with no Cloud Connector, and is not a full setup guide: the commands are left to the ARC-1 deployment docs.published 2026-04-29
ARC-1 on SAP BTP: Secure ABAP Agentic Development Beyond the Laptop
arc-1 has been through 3 major lines since this was published: 0.8 (2026-05), 0.9 (2026-05), 1 (2026-07). Affects A recipe.
Introducing ARC-1: A Secure ADT MCP Server for Enterprise SAP DevelopmentIntroducing ARC-1: A Secure ADT MCP Server for Enterprise SAP DevelopmentWhy ARC-1 exists beside earlier ADT MCP servers such as vibing-steampunk, fr0ster/mcp-abap-adt and dassian-adt: one centrally deployed instance per SAP system instead of a server on each developer laptop, with the MCP client to ARC-1 hop and the ARC-1 to SAP hop authenticated separately. It sets out the restrictive defaults (read-only, no free SQL or named table preview, writes confined to $TMP), the layering in which ARC-1's configuration, the user's role and SAP's own authorisation must all allow an action, Principal Propagation so SAP sees the real user, and audit events to the SAP Audit Log service; that is tool governance as much as agent access to your code. It leaves the BTP setup to the next part, and places ARC-1 beside SAP's ADT MCP Server, which SAP's configuration guide describes as a local HTTP server hosted by the IDE, and beside SAP API Policy 4.2026a, where ARC-1's stated position is to stay close to documented ADT behaviour and ask SAP before production use.published 2026-04-27
ABAP and Agentic AI: The Hidden Problem in Real ProjectsABAP and Agentic AI: The Hidden Problem in Real ProjectsAI in SAP development splits on three questions, how the model gets context, where writing happens and who controls the setup: UI5 and CAP work in local files where git, reviews, tests and pipelines already gate a change, while on-stack ABAP writes land directly in a shared development system. Six patterns follow, framed after a paper by Lars Hvam, from AI in the IDE to off-stack development with abapGit, and the post settles on a hybrid where the agent reads the system over ADT or MCP and writes through a gated git flow. Read access alone is put at maybe 80 percent of the work, and what agent access to your code still lacks is named as tool governance: read-only defaults, package restrictions, blocking free SQL, per-user identity and auditing.published 2026-04-22
How I Use AI for Development and Why Context MattersHow I Use AI for Development and Why Context MattersPart one of the AI ABAP development series traces a move from pasting ChatGPT output and error messages back and forth to tools that search, read docs and inspect files themselves, and holds that the context around a model decides more in real projects than benchmark results do. In SAP work the grounding comes from mcp-sap-docs, which brings SAP Help, GitHub repositories, style guides and DSAG guidelines into the session, and context for changing a class also means the tables it uses, the classes it calls and similar implementations. Agentic autonomy is a slider set per project, with the time saved on typing moved into planning and review, and the post stops before ABAP: SAP's MCP servers for UI5, CAP and Fiori (@ui5/mcp-server, @cap-js/mcp-server, @sap-ux/fiori-mcp-server) sit in off-stack work that git and pipelines already gate, and the harder ABAP case is left to the next part.published 2026-04-20
Building an Enterprise-Ready SAP AI Agent with Open SourceBuilding an Enterprise-Ready SAP AI Agent with Open SourceA self-hosted chat agent for SAP systems: LibreChat as UI and orchestrator with Entra ID sign-in and SharePoint connected, a local LLM through Ollama, and two MCP servers, mcp-sap-docs for public documentation and the ADT MCP server vibing-steampunk for the system itself. The ADT server runs in its read-only mode under a technical user with display authorisations only, so the agent can find a function module and pull its source, or locate a RAP business object with its behaviour definition and projection view, but cannot change anything. Two workflows show it, explaining what BAPI_DISPUTE_CREATE does in one system and planning a RAP action from a PDF specification, where the generated ABAP is copied into Eclipse by hand; the full setup is in the marianfoo/llm-client-sap-system-integration-sharepoint repository.published 2026-03-03
Finally: An MCP Server for ABAPFinally: An MCP Server for ABAPAnnounces the ABAP variant of mcp-sap-docs, published at mcp-abap.marianzeis.de for GitHub Copilot's Agent Mode in ADT for Eclipse, which the post names as the trigger, with ABAP knowledge still a recurring problem in the models. Documentation irrelevant to ABAP was removed from the SAP docs server and the ABAP Keyword Documentation, the DSAG ABAP Development Guidelines and the ABAP Style Guide were added, with search reaching past the local index to SAP Community posts and SAP Help. Copilot in Eclipse is where the post puts the weight, because SAP Joule for Developers, ABAP AI capabilities in Eclipse is not available for S/4HANA on-premise systems; an update of 2026-02-05 adds Software Heroes as an online source and an abap_feature_matrix tool that checks feature availability across releases and ABAP Cloud, and the server runs hosted or locally with Node or Docker.published 2026-02-04
IN TOPICS
onedevatlas is open for browsing. Find tools for SAP development, compare their documented capabilities, ask your AI assistant, and join a flight during TechEd week.
Details
History
- WerkbankWhat Werkbank adds for ABAP, CAP and UI5 is its catalog of development skills; the workspace itself is not specific to any of those stacks. Agent workspaces is narrowed back to BTP.
- WerkbankThe project homepage describes "A small, curated shelf of MCP servers and SAP development skills." Its own catalog spans ABAP (ARC-1 and ABAP Cloud), CAP (CAP Upgrade), and UI5 skills; the workspace is not specific to one stack. This intake follows the maintainer’s explicit request after the September source sweep passed it over for lacking a workspace capability; that concern now exists as Agent workspaces. The served views describe which users the project-owned install catalog addresses. Werkbank provides the workspace; each selected connector or skill supplies the stack-specific behavior. No SAP implementation capability is attributed to Werkbank itself.
- arc-1The licensing side is its explicit SAP API Policy relation, sourced to its README; the rest restated what any server in front of ADT is. Removed at the atlas maintainer's request on 2026-10-14.
- arc-1Re-read against the README, the npm registry record, the extensions page and the release notes on 2026-10-04, with arc-1 at 1.5.0 published 2026-10-02. Every claim the entry already carried still holds word for word, so the checked dates move and the text does not, except where the documentation says something the entry did not. Two such things: `Custom_*` tools an operator registers without forking, which the corpus already recorded on [claude-sap-kit](artifact:silvius1996-claude-sap-kit) and not on the server those tools run in, and one agent plugin that installs the server together with [ARC-1 Skills](artifact:arc-mcp-arc-1-skills). The seven guardrails added were stated when the entry was written and left out of it; they are quoted, not counted. The `run_code` facet stays absent: the extensions page gates class and report execution behind two flags and a plugin the operator writes, which is not the server stating that it runs code.
- arc-1The connection is what the README states the server fronts (docs/concepts/design-roles-and-interfaces.md, P1), and it attaches the system's entry. The two BTP connections are stated under the README's Cloud Foundry deployment heading, so they carry that condition; the Destination service's sentence states identity forwarding and no data flow, and the ABAP environment's sentence names no protocol. The README's credits table names abaplint as what the linter uses. Two older defects go with it: a nested link that rendered its outer bracket as text, and abap-adt-api named in the same quoted sentence as the other two inspirations and carried by no edge.
- abap-docsEntered as the repository behind two sentences the corpus already rests on. capability:documentation's abap framing says "The ABAP keyword documentation, the cheat sheets and the RAP tutorials are public repositories", and for the keyword documentation the public repository is this one, not SAP's: mcp-sap-docs carries it as the `abap-docs` submodule (https://github.com/marianfoo/abap-docs.git in .gitmodules), which is where its `abap-docs-standard` and `abap-docs-cloud` offline sources come from. doc-source rather than a provider, because that same capability's abap note states publishing a corpus is not answering the concern. identity.license is the `license` field of package.json; the GitHub API reports no licence for the repository, which is the second sentence of when_not_to_use.
- SAP API Policy Evidence SkillThe entry described the skill as gathering evidence "through MCP servers" without naming one, because four of the five were not in the corpus. Three entered in the same batch as their own packages, and mcp-sap-docs and arc-1 were already here, so the README's own prerequisites table becomes edges. Conditional rather than required: the README states the skill degrades gracefully when one is missing and names which two raise confidence to high.
- sap-note-search-mcpThe decline recorded for candidate:marianfoo/mcp-sap-notes on 2026-09-10 named this as the successor and as an open gap: "SAP Notes search is a real gap in the corpus and the thing that fills it has to be found some other way." Entered from the monorepo package rather than the archived repository, and from the package's own README.
- arc-1Corrected against the README, which documents a shared-credential mode (technical user) beside per-user identity.
- arc1-adt-abap-mcp-extThe cited sentence is no longer in the README: from 0.5.1 the plugin only contributes tools to SAP's supported server; waking the dormant one was 0.3.x.
- mcp-sap-docsCorrected against the README: its offline-sources table lists seventeen repositories, abap-docs-standard through wdi5.
- arc-1-lspTriaged out of data/crawled/mcp-candidates.json, the sap-ai-mcp-servers sweep. Written from the repository's own README; the catalogue is credited as where it was found and its words are not reused.
- arc1-adt-abap-mcp-extTriaged out of data/crawled/mcp-candidates.json, the sap-ai-mcp-servers sweep. Written from the repository's own README; the catalogue is credited as where it was found and its words are not reused.
- SAP API Policy Evidence SkillTriaged out of data/crawled/mcp-candidates.json, the sap-ai-mcp-servers sweep. Written from the repository's own README; the catalogue is credited as where it was found and its words are not reused. Hard rule 7: skills are indexed, never authored here.
- arc-1docs/NEIGHBOURS.md §2 assigns ARC-1 the posture ingest — as an entry — and §3.1 names its author, but nothing in data/artifacts held it. Entered as the artifact the posture calls for, and it is the first provider of capability:abap-agent-tooling in the corpus.
- arc-1abap-agent-tooling generalised into agent-code-access: the same concern exists in CAP, and the ABAP framing keeps the local name.
- mcp-sap-docsNamed in docs/NEIGHBOURS.md §3.1 as a project of the bestofcapjs author but never entered. It is a documentation retrieval server, not an index of options — the complement posture the same section assigns ARC-1, and hard rule 5 makes it exist here only as an attached entry.
- First entry seenDiscovered via GitHub API · arc-mcp, GitHub API · marianfoo, sap-ai-mcp-servers, GitHub organisations and accounts · marianfoo, Added by hand.
Profile
For your agent
- Ref
actor:marianfoo- JSON
- /actors/marianfoo.json
- MCP
- Connecting an agent