Everything in the atlas

Every gate here runs off-system, so what the class does on a real system and what the view does on the oldest UI5 release it has to serve are what the `pitfalls` catalogues carry and no local run can see. It runs abap2UI5 app classes, so ABAP that is not one has nothing here to deploy into. In system mode the sandbox tools are not offered, and an `app_act` event runs on the system as the configured user.

The tools are documented as meant for AI models and not a stable API, so anything built to parse `cds-mcp` output rather than to feed a model can break between releases. Its corpus is the CAP documentation, so Fiori elements, BTP service or ABAP questions are not in what `search_docs` indexes.

A descendant rather than a distinct idea: if what you need is the ADT surface itself, its ancestors are entered here too and one of them may already be in your stack.

It expects a real landscape, not just a project: a `.service.metadata` file naming the Mobile Services app, endpoint and destination, the Cloud Foundry CLI, and access to a BTP Cloud Foundry environment. From 0.4.0 macOS support is ARM64 only — Intel Macs are not supported — alongside Linux and Windows on x64 and ARM64.

It reads a UI5 project and the framework's own documentation, not a running system or a backend: a question about the service behind the app is answered somewhere else. Node.js v20.17 or v22.9 and above.

A Python 3.10+ CLI, installed with pipx rather than through any SAP-provided package manager. Its own README lists what it does not do yet: where-used, syntax check and unit test runs — for those it says to reach for [ADT/Eclipse](external:eclipse) or an MCP server. First release, v0.1.0, and no standalone LICENSE file in the repository — the MIT grant is in the README and in pyproject.toml's project metadata rather than its own file.

Commercial Its maker offers it commercially.

Functional boundary The MIT licence covers the connector only: connecting to SAP needs a licensed ABAPilot backend installed in your system, so this is a commercial decision wearing an open-source badge on one half. Its own published checks did not cover writes, activation, transports or authorisation boundaries.

A system that reports itself productive, or whose role the probe cannot prove non-productive, is write-locked with no override, so changing code in production is not something this does. The permission ladder constrains the package, object name and transport request the server itself passes on a write, and is not a sandbox around ABAP it executes: `abap_run`, `abap_test` and the opt-in `core.eval` run under the connected technical user's SAP authorisations, which is where the boundary actually sits. Data preview, dump variable contents and driving a transaction by batch input stay off in every mode, `admin` included, until each is named explicitly.

Not reachable Not reachable since 2026-10-20.

Functional boundary `run_class` executes code in your system — that is the point and it is also the risk.

Classic BOPF only: business objects of the ABAP Programming Model for SAP Fiori are not handled, and on [the BTP ABAP environment](artifact:btp-abap-environment) or S/4HANA Public Cloud classic BOPF enhancements do not exist. It is version 0.1.0, verified on S/4HANA 2025 (SAP_BASIS 816) on-premise only, and [arc-1](artifact:arc-mcp-arc-1) still marks the plugin API it is built on as experimental. ARC-1 loads no extensions on its multi-target routes and offers none of these tools in its `hyperfocused` tool mode.

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.

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.

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.

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.

Built and last touched in the same week of March 2026, so the maintenance reading is what it is — the authentication work is the reason to look, and a stalled repository is the reason to read it before depending on it.

Not for S/4HANA Public Cloud or [the BTP ABAP environment](artifact:btp-abap-environment), where the README states there is no `SUBMIT`, NAST or SOAP-RFC; the floor elsewhere is SAP_BASIS 7.50, the `/sap/bc/soap/rfc` service active, and an SAP user with `S_RFC` on the kit's function group as well as the authorizations for the reports and the spool. Print output is NAST messages only, so S/4 Output Management through BRF+ or APOC is outside it, and editing covers SmartForms alone, with Adobe Forms and SAPscript previewable but not editable. A report that opens a popup, calls a screen or is a dialog transaction cannot be run at all.

Freestyle SQL and `abap_run` mean an agent can read data and execute code, which is well past reading source. Neither is something to grant on a system where the distinction between development tooling and business data matters.

In the Eclipse mode the server lives in a developer's IDE, binds to 127.0.0.1 and answers any request carrying an `Origin` header with HTTP 403, so a pipeline or a remote agent has nothing to reach; the standalone mode on the runtime of the ADT for VS Code extension is the documented route with no IDE open. A new connection starts in write mode, so the 11 write tools are live against a connected system until someone presses Toggle Read-Only. The dropins route for the two bundles is documented as untested.

A compiled binary rather than an npm package, shipped prebuilt per platform and through pip and uvx. It collects anonymous usage telemetry by default, switched off with `--no-telemetry` or `DATAZOO_DISABLE_TELEMETRY=1`. And the BW reach is what distinguishes it — if the work is plain ABAP, that half is weight you are not using.

Two licences, by which package you install: the standalone server, `@mcp-abap-adt/core`, is AGPL-3.0-only, a licence to read before shipping anything built around it; the library, `@mcp-abap-adt/lib`, is Apache-2.0. The destination model also expects service keys sitting in a local directory, which is a secrets decision to make deliberately.

Development and sandbox systems only: the README warns against pointing it at production. Transport release is never executed, since the tool exists to show the policy stopping it, and separation of duties is checked against the `identity.approvers` list in the policy file rather than against SAP authorisations, so it holds as far as that file does and no further. The write path has been tested with ABAP classes and other object types may need adjustments, there is deliberately no tool that creates or deletes an object, and one server instance serves one SAP system.

Its own README opens by calling the server experimental. It also writes: locking, editing and activating in a system an agent can reach is a permissions decision before it is a tooling one.

Every tool its README lists reads — Get and Search, nothing that writes. Locking, editing, activation and transports are the other server, [mcp-abap-abap-adt-api](artifact:mario-andreschak-mcp-abap-adt-api), and mixing the two is how you end up with two MCP servers pointed at one system.

There is nothing to clone here: "Access is provided through a registration page. Once registered, you will receive access to the GitHub repository containing the agents, deployment guides, and documentation", and that repository answers 404 to anyone who has not registered, so neither the agent code nor the deployment guides can be read before the form is filled in. AWS states the agents are published as an unsupported solution, with customers responsible for testing and securing the code before deploying in production, and that they are intended for non-production SAP environments rather than production systems. The documentation link in the README is still a placeholder.

Real SAP calls need the proprietary [SAP NetWeaver RFC SDK](external:sap-nw-rfc-sdk) and the `pyrfc==3.3.1` release the README calls archived and yanked, on Python 3.11, so without those only the demo backend answers. Not over HTTP as it stands: credentials are read from the environment for stdio, and the README says not to expose the server over HTTP without implementing the MCP authorization specification and deployment-level rate limiting. Multi-tenant auth and RBAC beyond the one server-wide policy, rate limiting, metrics and tracing, and deployment packaging are listed as not built. The result limits protect the client and not the system, since the README states they cannot reduce work already performed inside SAP, and the MCP tool annotations that describe read and write risk are advisory metadata rather than an authorisation boundary, so the service account's `S_RFC` and table authorisations are still what decides.

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.

Read-only in practice, and the documentation is Korean; the tool names are English but the guidance around them is not.

Smart Form content cannot be read or written, and HR Forms (`PE51`) are not implemented. Installing it needs SAP GUI or WebGUI once, to register the OData services in `/IWFND/MAINT_SERVICE` and to redefine the write service in `SEGW`, and the extension is built against a locally linked ARC-1.

Every gate here runs off-system, so what the class does on a real system and what the view does on the oldest UI5 release it has to serve are what the `pitfalls` catalogues carry and no local run can see. It runs abap2UI5 app classes, so ABAP that is not one has nothing here to deploy into. In system mode the sandbox tools are not offered, and an `app_act` event runs on the system as the configured user.

It writes and it administers — ETL extraction and database user management are not read operations, and the same connection that profiles a column can move data out of the tenant. Masking is only as good as the configuration naming the fields, so an unlisted sensitive column is an unmasked one.

The embedded preview needs a system that accepts basic auth: the documentation states that pure SSO or SAML without a basic-auth fallback is not supported, and points at `external` mode, which opens the app in the normal browser with the existing SAP session. On the web only the language half runs; the preview with its proxy, traffic log and screenshot, Ctrl+F3 activation, the render gate, the workspace-wide check and the MCP servers are desktop-only. The agent app tools send nothing until `abap2ui5.agent.enableAppTools` is switched on in User settings, because an event they fire runs as the configured user and changes data on the system for real.

A system that reports itself productive, or whose role the probe cannot prove non-productive, is write-locked with no override, so changing code in production is not something this does. The permission ladder constrains the package, object name and transport request the server itself passes on a write, and is not a sandbox around ABAP it executes: `abap_run`, `abap_test` and the opt-in `core.eval` run under the connected technical user's SAP authorisations, which is where the boundary actually sits. Data preview, dump variable contents and driving a transaction by batch input stay off in every mode, `admin` included, until each is named explicitly.

Configuration over an API you must already have entitled and provisioned; and the smallest of these repositories, so what it covers is what its config file happens to name.

It administers a live Cloud Foundry landscape: `Apps`, `Organizations`, `Spaces`, `ServiceInstances` and `SecurityGroups` all carry create, update and delete alongside list and get. What a caller can reach is set by the role they hold, and the three templates in `xs-security.json` are MCPViewer for read, MCPEditor for read and write, and MCPAdmin for full administrative access.

It administers a real account through Core Services, so what it can change is what the caller can change. And the UI half depends on hosts that render mcp-ui or MCP Apps resources; elsewhere those five tools return markup nobody displays.

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.

A thin declaration over CPI's APIs: it reaches what those APIs expose and nothing else, and it inherits whatever the tenant's own authorisations allow.

Design-time only: no message monitoring, no deployment, no runtime status. That is a different server in this same section, and the two do not overlap.

This is the database, not an application in front of it, so what an agent can read is whatever the connection user can read — the guardrails bound the shape of a query, not the sensitivity of what it returns. Grant the connection the rights you would grant a person.

Not for production: the README states local and development instances only, and the server refuses a HAC URL whose host is not private unless `HAC_ALLOW_REMOTE=true` overrides that guard. Masking applies to `config_get` only, so `logs_tail` and `monitoring_info` output, thread dumps included, is unmasked and may carry sensitive data, and whatever a tool returns goes to the MCP client and its model. Certificate checks are disabled whenever `HAC_INSECURE_TLS=true` is set, which the README names as typical for local HAC.

It deploys and undeploys, and it enumerates credentials, keystores and OAuth2 configurations. Listing security material is a read, and it is still a read of the thing that protects everything else the tenant talks to.

It reads the landscape inventory, not the systems in it — nothing here reaches a running application, and nothing here changes anything. A landscape description is also commercially sensitive in its own right: versions and hosts are what a targeting exercise starts from.

OData v2 only, and untouched since August 2025 — the same author's [Go port](artifact:oisee-odata-mcp-go) is where v4 support and the tool-explosion work went.

Universal mode trades discoverability for size: one generic tool is harder for a model to pick correctly than a named one, which is why it is not the default. On a small service, standard mode is the better answer.

It fronts APIs you already have; nothing here reads a system's source or model, and a config file is only as good as the entity sets somebody wrote into it. OData V2 is the query surface it names.

Real SAP calls need the proprietary [SAP NetWeaver RFC SDK](external:sap-nw-rfc-sdk) and the `pyrfc==3.3.1` release the README calls archived and yanked, on Python 3.11, so without those only the demo backend answers. Not over HTTP as it stands: credentials are read from the environment for stdio, and the README says not to expose the server over HTTP without implementing the MCP authorization specification and deployment-level rate limiting. Multi-tenant auth and RBAC beyond the one server-wide policy, rate limiting, metrics and tracing, and deployment packaging are listed as not built. The result limits protect the client and not the system, since the README states they cannot reduce work already performed inside SAP, and the MCP tool annotations that describe read and write risk are advisory metadata rather than an authorisation boundary, so the service account's `S_RFC` and table authorisations are still what decides.

Ninety tools is a large surface to put in front of a model at once, and client-credentials OAuth means the tenant sees one identity rather than the person asking. Content transport and user management sit in the same tool list as reading a story.

On ECC the S/4HANA `API_*` services that the three convenience readers default to do not exist, so each one has to be pointed at a Z service of your own or replaced by the RFC tools. `read_table` inherits `RFC_READ_TABLE`'s 512-byte row, so a wide table needs an explicit field list and some fields cannot be selected at all, and the bridge starts a Node process and a fresh logon per call, with no pooling. The audit log is one line per call that includes the call arguments, so every filter and WHERE clause sent to SAP is written to a file that then has to be handled as sensitive.

Inspection only, and it takes a host, a user and a password from the environment rather than a managed credential. PI and PO are also on their own end-of-maintenance path, which is the context any investment here sits in.

Read-only does not mean low-stakes: a complete picture of who can do what in an SAP system is exactly what an attacker would want, and the ready-made endpoint is the vendor's rather than yours — the README also documents running the server yourself. The authentication is careful; the question of what leaves your landscape is still yours.

This is HR data — the most regulated business data most organisations hold — reached through a connection that sees whatever its principal sees. Employee records under natural-language query is a decision for whoever owns that data, not a configuration choice.

If your tool reads the open skill standard directly, install [capire/skills](artifact:capire-skills) and skip the wrapper.

Not a substitute for testing in a system: the README states that any class calling the database, classic BAPIs or HTTP is shimmed out and stays an on-stack ATC and ABAP Unit concern, and that an off-stack green is CI-grade provenance rather than an in-system pass. It also assumes a particular shape of project. The ABAP has to be abapGit-serialised into a Git repository, Node.js 22 or later has to be available locally and in CI, and the CI has to be GitHub Actions.

The embedded preview needs a system that accepts basic auth: the documentation states that pure SSO or SAML without a basic-auth fallback is not supported, and points at `external` mode, which opens the app in the normal browser with the existing SAP session. On the web only the language half runs; the preview with its proxy, traffic log and screenshot, Ctrl+F3 activation, the render gate, the workspace-wide check and the MCP servers are desktop-only. The agent app tools send nothing until `abap2ui5.agent.enableAppTools` is switched on in User settings, because an event they fire runs as the configured user and changes data on the system for real.

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.

Any run nobody is watching: the drawing tools work only while the draw.io canvas at `http://localhost:3000/` is open and connected, so there is no diagram to produce on a build agent or a server. Nothing checks the result either, because the maintainer keeps a diagram linter on the [roadmap](https://github.com/etosin/arc-draw#-roadmap) rather than in the release. The maintainer marks the status early at v0.x, installs it from source, and records testing with Claude Code alone.

Local scaffolding only. Its own Out of Scope list names deployment to Cloud Foundry or Kyma, XSUAA and IAS configuration or any other BTP service binding, the SaaS Registry and subscription flow to real tenants, and automatic migration from the legacy `@sap/cds-mtx`. For the concepts rather than the scaffolding step, the README points at the `multitenancy` skill in [skillOverflow](artifact:dankromp-skilloverflow) by [Daniel Krompaß](actor:dankromp), which covers the MTX sidecar, `xs-security.json`, tenant lifecycle hooks and the deployment path and runs no commands at all.

Knowledge with no reach: without an MCP server connected it can tell an agent what to write and cannot let it read or run anything. The rules target the ABAP Cloud development model, so a classic on-premise codebase is not what they are written against.

Not for S/4HANA Public Cloud or [the BTP ABAP environment](artifact:btp-abap-environment), where the README states there is no `SUBMIT`, NAST or SOAP-RFC; the floor elsewhere is SAP_BASIS 7.50, the `/sap/bc/soap/rfc` service active, and an SAP user with `S_RFC` on the kit's function group as well as the authorizations for the reports and the spool. Print output is NAST messages only, so S/4 Output Management through BRF+ or APOC is outside it, and editing covers SmartForms alone, with Adobe Forms and SAPscript previewable but not editable. A report that opens a popup, calls a screen or is a dialog transaction cannot be run at all.

The README tells you to fork before cloning, because Claude rewrites `references/` and `SKILL.md` in place as you use it, so a clone shared between people changes under all of them. Installing means `git` and `ln` on macOS or Linux and a symlink into `~/.claude/skills`. Fine positioning stays manual: the coordinates Claude writes can be rough apart from the centre coordinates of elements joined by a straight arrow and the containment relationships.

A requirement that needs code: the skill generates configuration and states that it says so instead of faking it. Dialog and container targets have no verified enter event, so a dashboard placed there loads only through its Refresh button. Production Processes are outside it — it wires PODs to them and does not design them.

It reaches a data warehouse, so what an agent can see is business data by definition — the SQL sanitisation and PII filtering in the server are there because of that, not instead of a decision about it. And the guidance is Datasphere-shaped: nothing here transfers to a HANA Cloud database used directly.

A diagram of anything other than an SAP, BTP or on-premise landscape: the skill is tuned to the Horizon style and its 71 templates. It drafts rather than finishes, and its README budgets 15 to 30 minutes of edits in draw.io per diagram. Without Python the validator does not run.

The commands it writes stop production systems, delete data and import transports, which is why execution is deliberately somebody else's decision. BTP and cloud operations are explicitly out of scope and planned as a separate plugin.

It stops before the artifact: in the README's own words it does not generate or deploy complete iFlow artifacts, so building and shipping the integration flows themselves stays with a person and a tenant. Knowledge beside the work rather than a generator, and it reads nothing from a running tenant.

Not for a project below the 2211-jdk21.1 baseline: older releases and the JSP Accelerator are covered only as explicitly marked legacy material, so a JSP storefront or a pre-JDK-21 platform is outside what the guides are written against. Knowledge with reach only through one script — it can tell an agent what to write and, apart from `query-items.sh`, cannot let it read what a given system contains.

Three of its five reference corpora are not shipped: without an api.sap.com key and session cookie, a tenant bearer token and a local SAPUI5 fetch, the REST, MDO and UI5 tools return a message naming the preparation step that is missing. The HTTP transport has no authentication until `MCP_AUTH_TOKEN` is set, so anything past localhost needs that or an authenticating reverse proxy. Packing the ZIP, uploading it through Manage POD 2.0 Extensions and placing the widget in the POD Designer stay manual steps outside the agent, and the SAPUI5 fetch defaults to the SDK's newest version, which the README says is not the version a tenant runs. Node.js 22 or newer.

Its README calls it early-stage — the core workflows run end-to-end against real systems and packaging, documentation and polish are stated as still evolving. A Rust binary rather than an npm install, so the platform build matters.

Skills that generate commands are not the commands: everything here still runs against a real BTP account with real entitlements, and the deployment path assumed is Cloud Foundry, with `cf login` and `joule login` done before the toolkit is used. Three of the four skills are CLI knowledge rather than anything Joule-specific, which is worth knowing if you only wanted the agent part.

It works against a tenant, through the REST API, with credentials in the environment — so these skills act rather than advise, and the review skill is a check on generated content rather than a substitute for one. Nothing here is useful without an Automation Pilot tenant to point at.

Enterprise-architecture work, not development: this describes what your landscape records, never what your code does. And the skills assume the server — without [the LeanIX MCP server](artifact:sap-leanix-mcp-server) reachable against your instance, they are instructions with nothing to call.

Knowledge about theming, not a theme build: nothing here compiles a theme or deploys one, and it says nothing about a custom control's own parameters.

It is a plugin for one server — its own — so it is a stack rather than a component. And the guard is only as good as the profile labels: a production system registered as `DEV` is a production system with the safety off.

Two skills, narrowly scoped: nothing here scaffolds an application or answers what a component's API is — that is [@ui5/webcomponents-mcp-server](artifact:ui5-webcomponents-mcp-server), which reads the same library's documentation.

Guidance, not access: these are instructions handed to an agent, and what the agent can then read of your project comes from [@ui5/mcp-server](artifact:ui5-mcp-server) or the editor. A modernisation workflow that runs autonomously across a codebase is also a thing to point at a branch rather than at main.