Supabase
MCP server for interacting with the Supabase platform
제작자가 쓴 설명(영어 원문)
Checks and fixes ABAP Open SQL for SAP's ADT data-preview console. Connects to nothing.
이 세 가지는 절대 하나의 점수로 합치지 않아요. 의견, 측정, 인기는 서로 다른 것이니까요.
이 서버는 아직 쉬운 말로 된 요약을 쓰지 못했어요. 아래는 레지스트리에서 읽어 온 사실이에요.
쓰고 있는 AI 앱을 고른 다음 설정을 복사하세요.
필요한 것
어떤 앱을 쓰세요?
위에서 앱을 고르세요. 기억해 두었다가 다음부터는 버튼에 “…에 설치”라고 표시해요.
Claude Desktop → Settings → Developer → Edit Config를 열고 claude_desktop_config.json에 이 내용을 붙여 넣으세요
{
"mcpServers": {
"sap-abap-sql": {
"command": "npx",
"args": [
"-y",
"sap-abap-sql-mcp"
]
}
}
}터미널에서 이 명령어를 실행하세요
claude mcp add sap-abap-sql -- npx -y sap-abap-sql-mcp프로젝트의 .cursor/mcp.json 또는 ~/.cursor/mcp.json에 붙여 넣으세요
{
"mcpServers": {
"sap-abap-sql": {
"command": "npx",
"args": [
"-y",
"sap-abap-sql-mcp"
]
}
}
}브라우저가 앱을 열지 물어본 다음, 앱이 설정을 보여 주고 확인을 요청해요. 키 값은 비워 두니 직접 입력하세요.
프로젝트의 .vscode/mcp.json에 붙여 넣으세요
{
"servers": {
"sap-abap-sql": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"sap-abap-sql-mcp"
]
}
}
}브라우저가 앱을 열지 물어본 다음, 앱이 설정을 보여 주고 확인을 요청해요. 키 값은 비워 두니 직접 입력하세요.
이 서버는 내 컴퓨터에서 실행되는데, ChatGPT는 인터넷에 있는 서버에만 연결할 수 있어요. 다른 앱을 고르세요.
선택한 앱은 이 기기의 쿠키에만 저장돼요. 계정은 필요 없어요.
이 설정은 제작자의 레지스트리 정보로 만들었어요. README와 다르면 README를 따르세요.
내 컴퓨터에서 실행돼요. 설치하면 AI 앱에 도구 목록이 나타나요.
누구나 볼 수 있는 정보를 점검했고, 코드는 읽지 않았어요. 해당하지 않는 항목은 개수에서 뺐어요. · 2026년 11월 6일 점검
레지스트리 이름이 도메인 daslab.run 기준으로 확인되었어요(레지스트리가 도메인 소유를 확인해요)
아직 GitHub에서 이 저장소를 읽어 오지 않았어요
아직 GitHub에서 이 저장소를 읽어 오지 않았어요
내 컴퓨터에서 실행되는 서버라 인터넷 엔드포인트가 없어요
내 컴퓨터에서 실행되며, 대신 로컬 설정을 사용해요
내 컴퓨터에서 실행돼요. 도구 목록은 인터넷 엔드포인트가 있는 서버에서만 읽을 수 있어요
레지스트리에서 이 카테고리의 활성 서버 전체와 비교했어요. 스타는 게시자가 저장소를 소유한 경우에만 세요.
해당하는 기본 점검 1개 중 1개를 통과했어요. 코드는 읽지 않았어요. 키나 토큰을 넣기 전에 서버가 어떤 권한을 요구하는지 확인하고, 꼭 필요한 권한만 있는 키를 쓰세요.
레지스트리에 필수 키가 나와 있지 않아요. 다만 처음 쓸 때 로그인을 요구하는 서버도 있어요.
내 컴퓨터에서 실행돼요. AI 앱이 필요할 때 실행하고, 데이터는 제작자의 서버를 거치지 않아요(서버 자체가 온라인 서비스를 호출하는 경우는 제외).
GitHub 저장소는 마지막으로 – 변경됐고, 최신 버전(0.3.0)은 3개월 전 레지스트리에 올라왔어요.
Claude Desktop, Claude Code, Cursor, VS Code처럼 MCP를 지원하는 앱에서 쓸 수 있어요. 내 컴퓨터에 Node.js도 있어야 해요.
같은 카테고리에는: Supabase, MongoDB MCP Server, Elasticsearch MCP Server
같은 카테고리에서 다른 제작자가 만든 서버예요. 스타 순으로 정렬했어요.
카드 보는 법:★ GitHub 스타i세 기둥: 리뷰 · 점검 · 사용 정도i업데이트 = 최신 버전을 레지스트리에 공개한 날i
MCP server for interacting with the Supabase platform
MongoDB Model Context Protocol Server
MCP server for interacting with Elasticsearch
제작자의 저장소에서 쓰인 그대로 가져왔어요. PopNx MCP가 검토하거나 보증한 내용이 아니에요. HTML, 이미지, 스크립트는 제거했어요. 링크를 누르면 이 사이트를 벗어나요.
Checks and fixes ABAP Open SQL before it reaches SAP's ADT data-preview console.
If you have pointed an LLM at an on-premise SAP system, you have seen these:
| You sent | SAP said |
|---|---|
SELECT SUM(netwr) FROM vbap | 400 Unknown column name "SUM(NETWR)" |
SELECT COUNT(DISTINCT kunnr) FROM vbak | 400 ")" is invalid here (due to grammar) |
SELECT a FROM t WHERE (x = '1' OR y = '2') | 400 ") " was expected here. |
Three errors, three apparent rules, one actual cause: the parser wants a space on the inside of every parenthesis. It even tells you in the third one. All three pass once spaced.
Nothing in those messages says "add spaces", so a model guesses. In one recorded session the same malformed statement went out ten times in a row, varying only a literal. Across two sessions, 57 of 211 tool calls failed, most of them on dialect.
This server knows the rules, so your agent does not have to rediscover them.
The ADT data-preview console (/sap/bc/adt/datapreview/freestyle), which is where agents and ADT clients read tables over HTTP. Open SQL written *inside an ABAP program* is a different dialect: it has INTO TABLE, UP TO n ROWS, @host variables, and no character ceiling. Do not point this at program source and expect it to be right.
On-premise systems (ECC, S/4HANA on-prem, private cloud). If you are on S/4HANA Cloud with proper OData APIs, you do not need any of this.
Only if you tell it to, and only from your own machine.
Three of the four tools are pure functions over a statement you pass in: no connection, no credential, no network call. They are the default, and they work with no configuration at all.
The fourth, abap_sql_query, runs a statement against a system you configure and returns rows. Credentials live in this server's environment on your machine. Nothing is proxied through anyone else and nothing is stored.
{
"mcpServers": { "sap-abap-sql": { "command": "npx", "args": ["-y", "sap-abap-sql-mcp"], "env": { "SAP_URL": "https://your-system:44300", "SAP_USER": "...", "SAP_PASSWORD": "...", "SAP_HEADERS": "X-Your-Proxy-Token: ..." } } } }
SAP_HEADERS is optional, one Name: value per line. On-premise systems are rarely reached directly: they sit behind a tunnel or proxy, and if that gates on a header of its own then every request is rejected before SAP ever sees it. The symptom is a 403 that reads exactly like "ADT is switched off", and is not.
Read-only by construction: the data-preview console cannot write, so there is no write path to misuse.
{
"mcpServers": { "sap-abap-sql": { "command": "npx", "args": ["-y", "sap-abap-sql-mcp"] } } }
| Tool | What it does |
|---|---|
abap_sql_prepare | Returns the statement to send, or refuses it and names the correction |
abap_sql_explain | Decodes a statement the console rejected, without rewriting it |
abap_sql_query | Runs a prepared statement against your configured system and returns rows |
abap_sql_rules | Lists the ways this dialect differs from standard SQL |
All read-only, no side effects.
abap_sql_prepare SELECT COUNT(DISTINCT kunnr) FROM vbak
→ { "ok": true, "statement": "SELECT COUNT( DISTINCT kunnr ) FROM vbak", "length": 40, "limit": 255 }
abap_sql_prepare SELECT a FROM t LIMIT 10 → { "ok": false, "rule": "no-limit", "fix": "ABAP Open SQL has no LIMIT. Cap the result with the reader's row limit instead." }
The five rules also arrive in the server's instructions on connect, so your client has them before it writes anything. abap_sql_rules is there for when you want them back explicitly.
Parenthesis spacing, everywhere it is required at once: aggregates, DISTINCT, subqueries, boolean groups. COUNT(*) stays compact, since the parser takes it either way and every character counts against the limit below.
The 255-character ceiling. The console accepts 255 and rejects 256, measured with whitespace collapsed. Wrapping across lines buys nothing, and the check runs *after* rewriting, because spacing makes statements longer. This one shapes how you plan a query: anything real has to be split and joined on your side.
`LIMIT`, `DESC`/`ASC`, and dot-qualified fields, refused with the ABAP form named. p.vbeln is not a typo the parser forgives, it is three tokens.
Where the literals are. A parenthesis inside 'PUMP (SPARE)' is data, and so is a double space. Both ABAP string forms and the comment form are understood, so a rewrite never changes what a statement asks for. This is the part a regex gets wrong quietly rather than loudly.
Two limits are the console's, not ours, and they shape everything you can ask:
255 characters per statement, so anything real has to be split into several statements and joined on your side. No window functions, no CTEs, no running totals.
5,000 rows, hard. A result at the cap is reported as truncated rather than handed back as if complete, because a client-side total over a silently truncated result is confidently wrong. That failure mode is worse than an error.
If you find yourself stitching four statements together to answer one question, that is the console, and no client fixes it.
This prepares statements and optionally runs them. It does not browse objects, execute ABAP, write anything, or know what your tables mean.
The dialect logic lives in abap-sql if you would rather call a library than run a server. Tokenizing is abaplint's lexer, which is the one correct ABAP tokenizer in TypeScript.
Built by Daslab. We connect agents to on-premise ERP systems. Connecting one from outside the network is its own problem, and there is a free write-up of what actually works: connecting an on-premise S/4HANA over HTTP.
git tag v0.3.0 && git push --tags. CI checks the tag against package.json, runs the tests, publishes to npm with provenance, builds the .mcpb bundle and attaches it to a GitHub release.
Publishing uses npm trusted publishing, so there is no token in this repo and nothing to rotate. npm trusts this repository and this workflow through OIDC.
Two steps stay manual because they need credentials CI does not hold: the MCP Registry (mcp-publisher publish) and Smithery (smithery mcp publish).
MIT