DISCLAIMER: This server is still in experimental status, use it with caution!

MCP-ADT

Description

This is a MCP-Server, it's a Model Context Protocol (MCP) server designed to facilitate seamless communication between ABAP systems and MCP clients. It is a wrapper for abap-adt-api and provides a suite of tools and resources for managing ABAP objects, handling transport requests, performing code analysis, and more, enhancing the efficiency and effectiveness of ABAP development workflows.

Features

• Authentication: Securely authenticate with ABAP systems using the login tool.
• Object Management: Create, read, update, and delete ABAP objects seamlessly.
• Transport Handling: Manage transport requests with tools like createTransport and transportInfo.
• Code Analysis: Perform syntax checks and retrieve code completion suggestions.
• Extensibility: Easily extend the server with additional tools and resources as needed.
• Session Management: Handle session caching and termination using dropSession and logout.
• Multiple Connections: Configure several named SAP systems (e.g. DEV, QAS, PRD) and switch between them at runtime with listConnections and switchConnection.

Installation

Prerequisites

• Node.js: Ensure you have Node.js installed. You can download it from here.
• ABAP System Access: Credentials and URL to access the ABAP system.

Steps

0. Allow commands in terminal if needed

   Set-ExecutionPolicy -ExecutionPolicy Bypass -Scope Process
   

1. Clone the Repository

   git clone https://github.com/rafaelssclaudio/MCP.git
   cd MCP
   

2. Install Dependencies

   npm install
   

3. Configure Environment Variables

   An .env.example file is provided in the root directory as a template for the required environment variables. To set up your environment:

   a. Copy the .env.example file and rename it to .env:

   cp .env.example .env
   

   b. Open the .env file and replace the placeholder values with your actual SAP connection details:

   SAP_URL=https://your-sap-server.com:44300
   SAP_USER=YOUR_SAP_USERNAME
   SAP_PASSWORD=YOUR_SAP_PASSWORD
   SAP_CLIENT=YOUR_SAP_CLIENT
   SAP_LANGUAGE=YOUR_SAP_LANGUAGE
   

   Note: The SAP_CLIENT and SAP_LANGUAGE variables are optional but recommended.

   Multiple connections (DEV, QAS, PRD, ...)

   The server can hold several SAP connections at once. Declare each one with a name suffix in the format SAP_<FIELD>_<NAME>:

   # Default active connection on startup (optional; first declared otherwise)
   SAP_DEFAULT_CONNECTION=DEV
   
   # --- DEV ---
   SAP_URL_DEV=https://172.16.6.60:44300
   SAP_USER_DEV=YOUR_USER
   SAP_PASSWORD_DEV=YOUR_PASSWORD
   SAP_CLIENT_DEV=100
   SAP_LANGUAGE_DEV=PT
   
   # --- QAS ---
   SAP_URL_QAS=https://172.16.6.60:44300
   SAP_USER_QAS=YOUR_USER
   SAP_PASSWORD_QAS=YOUR_PASSWORD
   SAP_CLIENT_QAS=110
   SAP_LANGUAGE_QAS=PT
   
   NODE_TLS_REJECT_UNAUTHORIZED="0"
   

   • A connection name is any suffix you choose (DEV, QAS, PRD, S4H, ...). The server discovers connections from every SAP_URL_<NAME> variable found.
   • URL, USER and PASSWORD are required per connection; CLIENT and LANGUAGE are optional.
   • The legacy unsuffixed format (SAP_URL, SAP_USER, ...) is still supported and is registered under the name DEFAULT. You can mix both.
   • The active connection defaults to SAP_DEFAULT_CONNECTION if set, otherwise the first declared connection (legacy DEFAULT wins if present, then names in alphabetical order).
   • Switch connections at runtime with the switchConnection tool and inspect the available ones with listConnections (see Custom Instruction below).
   • Each connection gets its own session/login; switching does not log the others out.

   If you're using self-signed certificates, you can also set:

   NODE_TLS_REJECT_UNAUTHORIZED="0"
   

   IMPORTANT: Never commit your .env file to version control. It's already included in .gitignore to prevent accidental commits.

4. Build the Project

   npm run build
   

5. Run the Server

   npm run start
   

6. Using it with Claude Code

   Install Claude-code in terminal

   npm install -g @anthropic-ai/claude-code
   

   Let Claude see the MCP ponting to the index.js from the dist where you installed it

   claude mcp add abap-adt --scope user -- node "C:\Users\Rafael\MCP\dist\index.js"
   

   (or alternatively integrate the MCP Server into VSCode)

Usage

Once the server is running, you can interact with it using MCP clients or tools that support the Model Context Protocol (e.g. Cline). In order to integrate the MCP Server with Cline, use the following MCP Configuration:

    "mcp-abap-abap-adt-api": {
      "command": "node",
      "args": [
        "PATH_TO_YOUR/mcp-abap-abap-adt-api/dist/index.js"
      ],
      "disabled": true,
      "autoApprove": [
      ]
    },

Custom Instruction fo AI use

Use this Custom Instruction to explain the tool to your LLM model:

## mcp-abap-abap-adt-api Server

This server provides tools for interacting with an SAP system via ADT (ABAP Development Tools) APIs. It allows you to retrieve information about ABAP objects, modify source code, and manage transports.

**Working with Multiple Connections:**

The server can be configured with several SAP connections (e.g. DEV, QAS, PRD). One connection is "active" at a time, and every other tool operates against it.

*   **`listConnections`:** Lists all configured connections and shows which one is active.
    *   No parameters.
    *   Returns `{ active, connections: [{ name, url, user, client, language, active }] }` (passwords are never returned).

*   **`switchConnection`:** Changes the active connection. All subsequent tool calls use it until switched again.
    *   `connection`: (string, required) The connection name to activate (e.g. `DEV` or `QAS`, as returned by `listConnections`).
    *   Each connection keeps its own session; you may need to `login` after switching to a connection you haven't used yet.

**Key Tools and Usage:**

*   **`searchObject`:** Finds ABAP objects based on a query string (e.g., class name).
    *   `query`: (string, required) The search term.
    *   Returns the object's URI.  Example: `/sap/bc/adt/oo/classes/zcl_invoice_xml_gen_model`

*   **`transportInfo`:** Retrieves transport information for a given object.
    *   `objSourceUrl`: (string, required) The object's URI (obtained from `searchObject`).
    *   Returns transport details, including the transport request number (`TRKORR` or `transportInfo.LOCKS.HEADER.TRKORR` in the JSON response).

*   **`lock`:** Locks an ABAP object for editing.
    *   `objectUrl`: (string, required) The object's URI.
    *   Returns a `lockHandle`, which is required for subsequent modifications.

*   **`unLock`:** Unlocks a previously locked ABAP object.
    *   `objectUrl`: (string, required) The object's URI.
    *   `lockHandle`: (string, required) The lock handle obtained from the `lock` operation.

*   **`setObjectSource`:** Modifies the source code of an ABAP object.
    *   `objectSourceUrl`: (string, required) The object's URI *with the suffix `/source/main`*.  Example: `/sap/bc/adt/oo/classes/zcl_invoice_xml_gen_model/source/main`
    *   `lockHandle`: (string, required) The lock handle obtained from the `lock` operation.
    *   `source`: (string, required) The complete, modified ABAP source code.
    *   `transport`: (string, optional) The transport request number.

*   **`syntaxCheckCode`:** Performs a syntax check on a given ABAP source code.
    *   `code`: (string, required) The ABAP source code to check.
    *   `url`: (string, optional) The URL of the object.
    *   `mainUrl`: (string, optional) The main URL.
    *   `mainProgram`: (string, optional) The main program.
    *   `version`: (string, optional) The version.
    *   Returns syntax check results, including any errors.

*   **`activate`:** Activates an ABAP object. (See notes below on activation/unlocking.)
    *    `object`: The object to be activated.

*   **`getObjectSource`:** Retrieves the source code of an ABAP object.
    *   `objectSourceUrl`: (string, required) The object's URI *with the suffix `/source/main`*.

**Workflow for Modifying ABAP Code:**

1.  **Find the object URI:** Use `searchObject`.
2.  **Read the original source code:** Use `getObjectSource` (with the `/source/main` suffix).
3.  **Clone and Modify the source code locally:** (e.g., `write_to_file` for creating a local copy, and using `read_file`, `replace_in_file` for modifying this local copy).
4.  **Get transport information:** Use `transportInfo`.
5.  **Lock the object:** Use `lock`.
6.  **Set the modified source code:** Use `setObjectSource` (with the `/source/main` suffix).
7.  **Perform a syntax check:** Use `syntaxCheckCode`.
8.  **Activate** the object, Use `activate`..
9.  **unLock the object:** Use `unLock`.

**Important Notes:**
*   **File Handling:** SAP is completly de-coupled from the local file system. Reading source code will only return the code as tool result - it has no effect on file. Files are not synchronized with SAP but merely a local copy for our reference. FYI: It's not strictly necessary for you to create local copies of source codes, as they have no effect on SAP, but it helps us track changes. 
*   **File Handling:** The local filenames you will use will not contain any paths, but only a filename! It's preferable to use a pattern like "[ObjectName].[ObjectType].abap". (e.g., SAPMV45A.prog.abap for a ABAP Program SAPMV45A, CL_IXML.clas.abap for a Class CL_IXML)
*   **URL Suffix:**  Remember to add `/source/main` to the object URI when using `setObjectSource` and `getObjectSource`.
*   **Transport Request:** Obtain the transport request number (e.g., from `transportInfo` or from the user) and include it in relevant operations.
*   **Lock Handle:**  The `lockHandle` obtained from the `lock` operation is crucial for `setObjectSource` and `unLock`. Ensure you are using a valid `lockHandle`. If a lock fails, you may need to re-acquire the lock. Locks can expire or be released by other users.
*   **Activation/Unlocking Order:** The exact order of `activate` and `unLock` operations might need clarification. Refer to the tool descriptions or ask the user. It appears `activate` can be used without unlocking first.
* **Error Handling:** The tools return JSON responses. Check for error messages within these responses.

## Efficient Database Access

SAP systems contain vast amounts of data.  It's crucial to write ABAP code that accesses the database efficiently to minimize performance impact and network traffic.  Avoid selecting entire tables or using broad `WHERE` clauses when you only need specific data.

*   **Use `WHERE` clauses:** Always use `WHERE` clauses in your `SELECT` statements to filter the data retrieved from the database.  Select only the specific rows you need.
*   **`UP TO 1 ROWS`:** If you only need a single record, use the `SELECT SINGLE` statement, if you can guarantee that you can provide ALL the key fields for the `SELECT SINGLE` statement. Otherwise, use the `SELECT` statement with the `UP TO 1 ROWS` addition. This tells the database to stop searching after finding the first matching record, improving performance. Example:

    ```abap
    SELECT vgbel FROM vbrp WHERE vbeln = @me->lv_vbeln INTO @DATA(lv_vgbel) UP TO 1 ROWS.
      EXIT. " Exit any loop after this.
    ENDSELECT.
    ```
## Checking Table and Structure Definitions

When working with ABAP objects, you may encounter errors related to unknown field names or incorrect table usage.  You can use the following tools to inspect table and structure definitions:

*   **`GetTable`:** Use this tool to retrieve the structure of an ABAP Dictionary table, including its field names and data types. This is helpful for verifying the correct fields to use in your `SELECT` statements.
*    If you need to inspect an include structure, you may need to use `searchObject` to find the include and then use `GetTypeInfo` or `GetStructure`. You may get a 404 error and try again with `GetStructure`
*   **`GetStructure`:** Use this tool to retrieve the structure of an ABAP Dictionary structure, including its field names and data types. This is helpful for verifying the correct fields to use in your `SELECT` statements.
*    If you need to inspect an include structure, you may need to use `searchObject` to find the include and then use `GetTypeInfo` or `GetStructure`.