Create a notebook for SAP's VS Code extension¶

Ask your agent to save a native .abapsql file, then open it with ABAP Development Tools for VS Code. SAP's extension handles editing, syntax checks and execution. ARC-1 can help discover table fields and try individual queries through SAPQuery.

Start with an example¶

Copy the flight example to your workspace. It uses SAP's /DMO demo tables; adapt it if those tables are absent.

Open the .abapsql file in VS Code. If it opens as text, use Reopen Editor With… → ABAP SQL Notebook. To start empty, run ABAP: SQL Notebook.
Select each SQL cell and run ABAP: Set Cell System. The example deliberately has no system binding. Choose your existing destination; do not invent one.
Review the SQL, then use ABAP: Set Cell Max. Rows and ABAP: Check Cell Syntax. A row cap limits returned rows, not necessarily database work.
Run the intended cells explicitly and save the file. Opening it does not execute SQL.

Execution uses the extension's own destination and SAP identity. SAP authorization still applies; ARC-1's SQL/data gates, blocklist, deny rules, scopes and audit do not. ARC-1 controls apply only when a query is sent through ARC-1.

Ask an agent to create one¶

Create a native SAP .abapsql notebook for [question]. Verify table and field names with the available ARC-1 tools. Put the purpose and result interpretation in Markdown cells, followed by focused ABAP SQL cells with rowNumber: 20. Omit destination so I can select the system in VS Code. Save the file without executing it. Use a separate aggregate query when the question needs a count over all matching records; do not infer that count from a capped detail result.

The file is ordinary UTF-8 JSON. This is the format observed in SAP extension 1.2.0:

{
  "cells": [
    { "kind": "markdown", "value": "# Flight count\nCount the matching flights." },
    { "kind": "sql", "value": "SELECT COUNT( * ) AS N FROM /DMO/FLIGHT", "rowNumber": 20 }
  ]
}

Field	Meaning
`cells`	Ordered Markdown and SQL cells
`kind`	`markdown` or `sql`
`value`	Markdown or ABAP SQL text
`destination`	Optional exact extension destination name; set through the editor
`rowNumber`	Optional integer row cap; extension default 100, editor range 1–100,000

The extension saves these fields, not arbitrary custom metadata, assertions or result rows. Keep durable notes in Markdown cells. This observed serialization is not a versioned SAP interchange specification; recheck it after extension updates. No ARC-1-specific schemaVersion, cell IDs, target mapping or import step is needed. Validate the JSON before opening: malformed JSON opens as an empty SQL cell in 1.2.0, and saving that state replaces the original contents. The supplied example already matches the serializer's formatting. ARC-1's separate maxRows cap is 10,000.

Interpret query results¶

Use ABAP SQL (alias~field, DESCENDING), and a row cap rather than SQL LIMIT. See SAPQuery syntax and backend limits.

For ARC-1 query responses, rowsReturned counts output rows. SAP's totalRows can describe matching input records. Observed on SAP_BASIS 758: a scalar COUNT returns one row containing "0 " while totalRows is 0. Read the aggregate value from its result column. Values remain strings, including padding, leading zeros and decimal precision. Separate cell executions are not an atomic snapshot.

SAP extension · Extraction decision and validation