This paper presents the architecture of embedded real-time web server. Unlike existing web servers, in our approach, requests are processed not in the “first in first out” order but according to their deadlines and the expected server load. For this purpose the Least Laxity First scheduling method is used. First, requests with imposed hard real-time constraints are served. Then requests enclosed by soft deadlines are processed. Finally, request without time requirements are served in the order they arrived. We also present real-time extensions to the Hypertext Transfer Protocol. We propose headers that enable defining hard and soft deadlines, as well as responses containing time information, that are being sent to the client application. The experimental results showed that in case of real-time applications our server misses significantly fewer requests, due to time out, then existing solutions. The presented server may be very useful for implementing real-time services supported by embedded systems, e.g. in future real-time “Internet of things” applications.