Our service offers great reading accuracy but it won't work in certain cases. As a rule of thumb if a human has a hard time reading the license plate, our service will not be able to read it. If a human is able to read the license plate, we will likely read it too. The main use case for our service is automation.
We currently do not support on premises plate recognition. You have to call our cloud API to use our service.
You can find the documentation and examples of code on our API documentation page.
Depending on your use case you may have good results with an open source OCR. I believe it should work well if the images are very clean. By clean I mean good contrast, good resolution, sharp image and not a lot of deformation (you take the picture right in front of the license plate). That approach will struggle if your images has some of those features: