Multi-Agent systems facilitate the process of accessing, processing and gathering 
information on the Internet. Developing these systems from scratch comprises 
implementation of protocols, name services and other features similar for all 
types of agents. Agent building tools offer class libraries for these common 
functions. This allows the developer to concentrate more on the application 
specific issues. There exists a number of toolkits that offer different levels 
of design support, which makes it difficult to the developer to select the 
appropriate one. This report evaluates various toolkits regarding their 
suitability to implement multi-agent systems for information gathering. 
In addition, a benchmark for agent building tools is introduced. Finally, 
a prototype for an information gathering system is implemented using a toolkit 
selected in the evaluation process. This prototype consists of two types of agents, 
namely interface and resource. The interface agents interact with users, while 
the resource agents access, retrieve and manipulate the requested information. 
Three search engines (AltaVista, Google and Excite) are used as meta data 
by three different resource agents.