GOLEMlite is a framework built to facilitate the development of Java-based multi-agent systems, and demonstrate how its features benefit researchers, as well as students new to the area of AI agents. By providing a Java library with clearly documented concepts, simple APIs and carefully naming classes and methods, it allows students to focus on the AI concepts they are learning, instead of spending their time grokking the framework.

The GOLEMlite framework can be used to develop Java agents or Java/Prolog agents, and the environment in which they are situated. It provides a set of well de fined classes which an be extended or included to build a multi-agent system. It attempts to abstract away from the application developer low-level constructs such as agent architecture (how the cycle-step is implemented) or the underpinnings of how communication occurs by use of base classes which take care of these details.


To work on the framework itself, install Maven (3.0.5 was used for this project), and Git. Java 7 is required, and building using Maven on the command line will require an environment variable named "JAVA_1_7_HOME" (as per the pom.xml file) pointing to the home directory of your Java 7 installation.

To use the classes in the framework (e.g. to create a multi-agent system, or GOLEMlite app) create a jar that can be added to your project by executing

mvn jar:jar

which will compile the project and assemble the jar in the target directory. See the Maven documentation for more information.


Access to the GOLEMlite project is available via the BitBucket repository linked, either by cloning the repository locally, or downloading a zipped version of the project here.


GOLEMlite is a Java library, built using the concepts upon which GOLEM was created. It was developed primarily by past-DICE lab member, Ataul Munim, as part of his Master's degree.


The work on GOLEMlite was funded by Royal Holloway, University of London.


Agents, Java, Prolog, Single Container

View on bitbucket