The idea of an API-driven OCR service came up at last month's iDigBio Augmenting OCR Hackathon.

I wasn't involved in the team that built it, as I got distracted detecting handwritten sources from OCR output, so I'm afraid I don't know very much about how far they got.  Nevertheless, I'd recommend taking a look at the documentations for the REST API they developed:

https://github.com/idigbio-aocr/RESTAPI/tree/master/doc

Ben Brumfield
http://manuscripttranscription.blogspot.com/