Our study investigates political candidates' networks in the multilevel political setting of Belgium. Using Twitter data collected during the four months preceding the May 2019 regional, federal and European elections, we examine the extent to which network homophily - defined as the tendency to interact with similar others - occurs among political candidates along parliament, language and party lines. Relying on a unique dataset of 20 061 retweets between 935 candidates, we find that network interactions are most likely to occur among co-partisans and candidates speaking the same language. Candidates campaigning for the same parliament also tend to retweet among each other, although this tendency is not strong. Overall, the findings confirm the strong divide of Belgian politics along language lines and Belgium's 'partitocracy' in which parties are the main actors in the representation and policy-making process.