There is a Stat Forecasting Engine app on the App hub (a monthly and a weekly version). There are 30 different forecasting methods built in to the app. I can't see ARIMA specifically, but there are regression and moving average methods. It should be possible to take some of the modelling techniques used for these methods and build out ARIMA. Attached is a file with the included methods listed.
Hope that is of interest. I'd be interested to learn more of your use case and follow your journey to solve this.
In addition to that application that @andrew_martin_1 mentioned you can implement the same logic yourself without too much difficulty.
I would suggest this anyway because with most ARIMA models, there are a number of nuances that you need to address that you won't find in the generic application.
The use case I use most often in the retail industry is what's called VRS, or variable response smoothing. It's a adaptive exponential smoothing forecast. It uses last year's seasonal pattern and applies the current "trend" to that pattern to predict the future (that's the ARIMA part).
Trend Signal = Mean Signed Error / Mean Absolute Error.
Here's an example: You can see the forecast "accelerates" to chase the trend then as the forecast becomes more accurate it slows down. Nice inflection point, no?