Introduction to MCP Integration Suite
With this blog I would like to showcase how you can extend Integration Suite development with AI.
To extend the capability of AI with other tools we can use AI Chat clients along with the ModelContextProtocol (MCP). If you want to know more about this I recommend you this blog by @MarioAndreschak
I created a custom MCP Server to give LLMs access to Integration Suits content and lets it perform analysis or development operations.
What Can You Do With It?
The MCP Integration Suite Server enables you to perform key integration tasks through AI assistants:
Package Management – Create, retrieve, and manage integration packagesIntegration Flow Management – Create, modify, deploy, and visualize IFlowsMessage Mapping – Create and update message mappings and associated data typesTesting – Send HTTP requests to integration endpointsMonitoring – Analyse Messages from the monitoring tab
Setup and Configuration
I highly recommend using the tool with Cline, which is an extension for VSCode, but you can use every client that supports MCP. The installation is also mentioned in @MarioAndreschak ‘s blog above. For LLM providers I had the best experience with Google’s Gemini 2.5 pro.
You can get an API Key under https://aistudio.google.com/app/apikey You will probably have to create a billing account in the Google Cloud Console, but currently they offer up to two free requests per minute, which was enough for me.
Another LLM option is Claude which costed like 0.50$ per task for me.
Get your API Key and add it to claude. Next you want to install the Integration Suite MCP Server in Cline.
Get started by cloning the project and setting up authentication. Make sure you have NodeJS > 20 and NPM installed on your machine
git clone https://github.com/1nbuc/mcp-integration-suite.git
cd mcp-integration-suite
npm install
npm run build
cp .env.example .env
Next step is to set up Authentication and Links to connect with IS API.
You can either use basic authentication with your S-User credentials
Option 1:
For the API_BASE_URL please try to use another Integration Suite API instance in your subaccount. I didn’t find a way to get this URL otherwise as of right now. So use option 2 if this doesn’t work for you
API_BASE_URL=https://<cpi uri>.<tenant>.hana.ondemand.com/api/v1
# Credentials from https://account.sap.com/manage/accounts user account, not your universal ID
API_USER=S001234567
API_PASS=ooOoPsieEPWLeak-_-
Or create OAuth2 Client/Credentials in BTP by creating Service keys on an API Instance:
Option 2:
Assign the permissions you want the AI to have:
Click on the Created Instance and Create a Service Key:
View the service key and add the corresponding values in the projects .env
For the API_BASE_URL append the URL from the service key with /api/v1
Creating credentials for Sending messages (optional):
This is only used for sending messages against an IFlow to test it, which can be quite useful.
You can create another instance/service key the same way but on the instance creation popup use integration-flow instead of api. For the CPI_BASE_URL use the URL you send data to without any paths
Then in the .env add corresponding values to
# only used for sending test messages to IS
CPI_BASE_URL=https://<cpi url>.<tenant>.hana.ondemand.com
CPI_OAUTH_CLIENT_ID=
CPI_OAUTH_CLIENT_SECRET=
CPI_OAUTH_TOKEN_URL=https://<subaccount>.authentication.<tenant>.hana.ondemand.com/oauth/token
Next, add the MCP Server in Cline:
Replace the JSON with this. Change the project Path to your local project Path
{
“mcpServers”: {
“mcp-integration-suite”: {
“autoApprove”: [],
“disabled”: false,
“timeout”: 300,
“command”: “node”,
“args”: [
“<project Path>/dist/index.js”
],
“transportType”: “stdio”
}
}
}
After saving you should see the server in the list
What can you actually do with it?
Here are some examples that I created using the above configuration that worked, without any other instructions.
Example 1: Creating a Basic HTTP IFlow
“Create a new iflow in package ‘MyPackage’ with name ‘if_simple_http’ that receives data via HTTP on /simplehttp and sends it to https://echo.free.beeceptor.com via HTTP Post.”
The tool will generate all necessary configuration and can even deploy the IFlow for you.
You can also specify weather the IFlow should be deployed or not. If the deployment fails, the AI will try to fix the IFlow based on the deployment error.
Example 2: Message Mapping Development
Creating message mappings between complex structures is significantly simplified:
“Create a message mapping called ‘mm_invoice_demo’ in package ‘MyPackage’ which maps between two different invoice datatypes. Generate the datatypes as well”
The AI will generate both the mapping and the required XSD schemas.
Generate XML Data based on the source type of mm_invoice_demo
Example 3: Monitoring Integration Messages
Need to check for errors? Just ask:
“Get all messages with errors from the last two days.”
The tool will fetch and display the relevant error messages, making troubleshooting much faster.
You can also run more complex queries like “How much messages were processed this year per iflow accross all packages”
How to improve results:
The tool does pretty well on read-only stuff like describing complex scenarios or querying message logs. It has its struggles when you try to develop more complex IFlows or mappings, a good way to resolve these issues is giving existing examples like “Generate a iflow which …instructions… It is very similar to iflow if_very_similar_iflow”
Conclusion
The MCP Integration Suite Server can significantly improve productivity when working with SAP Integration Suite. Especially doing read only stuff like message monitoring almost always works without errors. What I personally like to do a lot is generating test data with a promp like “generate XML data for source type of mapping XY”
Let me know your thoughts/experiences on this topic in the comments!
Also you can check out the project readme for more info
Introduction to MCP Integration SuiteWith this blog I would like to showcase how you can extend Integration Suite development with AI.To extend the capability of AI with other tools we can use AI Chat clients along with the ModelContextProtocol (MCP). If you want to know more about this I recommend you this blog by @MarioAndreschak I created a custom MCP Server to give LLMs access to Integration Suits content and lets it perform analysis or development operations. What Can You Do With It?The MCP Integration Suite Server enables you to perform key integration tasks through AI assistants:Package Management – Create, retrieve, and manage integration packagesIntegration Flow Management – Create, modify, deploy, and visualize IFlowsMessage Mapping – Create and update message mappings and associated data typesTesting – Send HTTP requests to integration endpointsMonitoring – Analyse Messages from the monitoring tab Setup and ConfigurationI highly recommend using the tool with Cline, which is an extension for VSCode, but you can use every client that supports MCP. The installation is also mentioned in @MarioAndreschak ‘s blog above. For LLM providers I had the best experience with Google’s Gemini 2.5 pro.You can get an API Key under https://aistudio.google.com/app/apikey You will probably have to create a billing account in the Google Cloud Console, but currently they offer up to two free requests per minute, which was enough for me.Another LLM option is Claude which costed like 0.50$ per task for me.Get your API Key and add it to claude. Next you want to install the Integration Suite MCP Server in Cline.Get started by cloning the project and setting up authentication. Make sure you have NodeJS > 20 and NPM installed on your machinegit clone https://github.com/1nbuc/mcp-integration-suite.git
cd mcp-integration-suite
npm install
npm run build
cp .env.example .envNext step is to set up Authentication and Links to connect with IS API.You can either use basic authentication with your S-User credentials Option 1:For the API_BASE_URL please try to use another Integration Suite API instance in your subaccount. I didn’t find a way to get this URL otherwise as of right now. So use option 2 if this doesn’t work for youAPI_BASE_URL=https://<cpi uri>.<tenant>.hana.ondemand.com/api/v1
# Credentials from https://account.sap.com/manage/accounts user account, not your universal ID
API_USER=S001234567
API_PASS=ooOoPsieEPWLeak-_-Or create OAuth2 Client/Credentials in BTP by creating Service keys on an API Instance:Option 2:Assign the permissions you want the AI to have:Click on the Created Instance and Create a Service Key:View the service key and add the corresponding values in the projects .envFor the API_BASE_URL append the URL from the service key with /api/v1Creating credentials for Sending messages (optional):This is only used for sending messages against an IFlow to test it, which can be quite useful.You can create another instance/service key the same way but on the instance creation popup use integration-flow instead of api. For the CPI_BASE_URL use the URL you send data to without any paths Then in the .env add corresponding values to# only used for sending test messages to IS
CPI_BASE_URL=https://<cpi url>.<tenant>.hana.ondemand.com
CPI_OAUTH_CLIENT_ID=
CPI_OAUTH_CLIENT_SECRET=
CPI_OAUTH_TOKEN_URL=https://<subaccount>.authentication.<tenant>.hana.ondemand.com/oauth/tokenNext, add the MCP Server in Cline:Replace the JSON with this. Change the project Path to your local project Path{
“mcpServers”: {
“mcp-integration-suite”: {
“autoApprove”: [],
“disabled”: false,
“timeout”: 300,
“command”: “node”,
“args”: [
“<project Path>/dist/index.js”
],
“transportType”: “stdio”
}
}
}After saving you should see the server in the list What can you actually do with it?Here are some examples that I created using the above configuration that worked, without any other instructions.Example 1: Creating a Basic HTTP IFlow”Create a new iflow in package ‘MyPackage’ with name ‘if_simple_http’ that receives data via HTTP on /simplehttp and sends it to https://echo.free.beeceptor.com via HTTP Post.”The tool will generate all necessary configuration and can even deploy the IFlow for you.You can also specify weather the IFlow should be deployed or not. If the deployment fails, the AI will try to fix the IFlow based on the deployment error.Example 2: Message Mapping DevelopmentCreating message mappings between complex structures is significantly simplified:”Create a message mapping called ‘mm_invoice_demo’ in package ‘MyPackage’ which maps between two different invoice datatypes. Generate the datatypes as well”The AI will generate both the mapping and the required XSD schemas.Generate XML Data based on the source type of mm_invoice_demoExample 3: Monitoring Integration MessagesNeed to check for errors? Just ask:”Get all messages with errors from the last two days.”The tool will fetch and display the relevant error messages, making troubleshooting much faster.You can also run more complex queries like “How much messages were processed this year per iflow accross all packages” How to improve results:The tool does pretty well on read-only stuff like describing complex scenarios or querying message logs. It has its struggles when you try to develop more complex IFlows or mappings, a good way to resolve these issues is giving existing examples like “Generate a iflow which …instructions… It is very similar to iflow if_very_similar_iflow” ConclusionThe MCP Integration Suite Server can significantly improve productivity when working with SAP Integration Suite. Especially doing read only stuff like message monitoring almost always works without errors. What I personally like to do a lot is generating test data with a promp like “generate XML data for source type of mapping XY”Let me know your thoughts/experiences on this topic in the comments!Also you can check out the project readme for more info Read More Technology Blogs by Members articles
#SAP
#SAPTechnologyblog
