Setup

Setup

0. Prerequisites

• Node.js (https://nodejs.org)
• git
• AI IDE with MCP support
• SAP system to connect to
• Windows 10+ x64 / - linux/amd64 (expected to work on modern glibc-based distributions as Ubuntu 22.04+, Debian 12+, RHEL...)

1. Registering the Bridge

Note: Every AI IDE slightly differs in how it handles MCP server and skill registration. Read the documentation of your AI IDE or ask your AI agent to help you with the registration process.

The sap-bridge operates as an orchestrator-agnostic MCP server, fully supporting modern AI IDEs and clients.

It uses standard process stdio pipelines to communicate with your AI securely and silently in the background.

To register the bridge into your orchestrator's MCP config file, assign the absolute path to sap-bridge (found in bin folder) or follow your AI IDE's documentation.

Windows

{
  "mcpServers": {
    "sap-bridge": {
      "command": "C:\\absolute\\path\\to\\sap-bridge.exe",
      "args": [],
      "env": {}
    }
  }
}

Linux

{
  "mcpServers": {
    "sap-bridge": {
      "command": "/absolute/path/to/sap-bridge",
      "args": [],
      "env": {}
    }
  }
}

macOS

{
  "mcpServers": {
    "sap-bridge": {
      "command": "/absolute/path/to/sap-bridge-darwin-[arm64|amd64]",
      "args": [],
      "env": {}
    }
  }
}

2. Setting Up Credentials

1. Ask your AI Agent for the current URL of the dashboard (it will run sap_bridge_status MCP tool and return http://127.0.0.1:<some_local_port>).
2. Open the provided URL in your standard Web Browser.
3. If this is your first time securely booting the Bridge, the Dashboard will check the prerequisites and automatically prompt you to establish your Master Password. This will generate .sap-bridge-dashboard.json in your user home directory as well as .sap_credentials.enc in the project folder.
4. Once your Web UI encryption is initialized, you can add SAP system connections.

• You can toggle system visibility (to the agent) and writability on the fly.
• Additional controls will be shown depending on the chosen connection type and whether the password should be injected at runtime.

3. Creating the Agent Sandbox

To enable the AI Agent's isolated syntax checking mechanism, an empty ABAP report named Z_AGENT_SANDBOX must be created in the target SAP system.

1. Open SAP GUI or Eclipse ADT connected to your target system.
2. Navigate to package $TMP (Local Objects).
3. Create a new ABAP Executable Program (Report) named exactly Z_AGENT_SANDBOX.
4. Leave it entirely empty and save/activate it. No further configuration is required.

4. Setting Up RPC Proxy (Optional)

Some tools (like sap_read_screen) cannot reuse an existing adt endpoint and rely on ABAP logic. To make them work, you must deploy the ABAP RPC proxy class to your SAP system and configure an ICF node.

1. Deploy Proxy Class: Have the Agent push the offline ZCL_SAP_DEV_RPC class to your SAP system (via sap_push_source), or create the class manually using the raw ABAP code located in references/zcl_sap_dev_rpc.clas.abap. Activate it.
2. Create ICF Node:
   • Open SAP GUI and run transaction SICF.
   • Navigate to /default_host/sap/bc.
   • Create a new sub-element named zsapdevrpc (or any custom path).
   • In the Handler List tab, enter ZCL_SAP_DEV_RPC.
   • Save and Activate the ICF node.
3. Enable in Dashboard: Open the SAP-Bridge Web Dashboard, edit your connection, click the Agent Settings icon (file with a gear), and check the RPC Proxy Enabled checkbox. If you used a custom ICF path, specify it in the Agent RPC Path field (defaults to /sap/bc/zsapdevrpc).

Registering the Skill

Place the contents of sap-dev folder into your AI IDE's skill directory.