The Wumpus World is a didactic maze game providing the simplest possible intelligent agent environment that is still challenging enough to illustrate many key AI concepts. The goal of the simulator is to serve an the first building block towards an intelligent agent verification controlled experiment workbench able to check whether a given agent sensing, reasoning and acting in the Wumpus World satisfies some properties such as its ability to reach its goal safely in one specific or all instances of the world.

The simulator is a web service accepting http requests from a client agent sending as payload exploration actions to execute in the Wumpus World. As a response it returns a JSON object containing the percepts of the agent in the new state of the world updated by the consequences of the sent action, together with some additional simulation status information.

It is implemented as a Constraint Object-Oriented Logic Program and is based on Michael Thielscher's fluent calculus for the logical formalization of state change.

See the github repository for details on use and installation. See also https://youtu.be/ZHDLgZ-3FBQ for a short demonstration of the tool in action.