For the semantic indexing of Arabic documents, we are building a new ontology (SchemNet) based on Arabic schemes and their meanings and the combination of the contents of Arabic WordNet (AWN) and Arabic VerbNet (AVN). Then we use this ontology to filter documents in order to find the relevant documents that respond to the user requests. Although the indexing process is based on the stemming phase, the current available solutions have proven that this step consumes a lot of time when processing a large volume of data. To the best of our knowledge, there are no works addressing this problem and especially through the use of agents to improve the indexing process of Arabic documents. Our objective is indexing Arabic documents based on the multi-agent system to improve the indexing process on the term of time processing.