This document proposes a novel Software Agent Based Forest Fire Detection (SAFFD) approach that uses software agents to disseminate data and make decisions about forest fire detection in a wireless sensor network. The SAFFD approach works in two phases: 1) a data collection phase where a mobile agent collects sensed data from sensor nodes and aggregates it, and 2) a risk analysis phase where an actor node analyzes the aggregated data to determine the fire risk level (green, yellow, or red) and take appropriate action. The implementation of SAFFD using Aglets mobile agents showed significant increases in the lifetime of the wireless sensor network by reducing the number of transmitted packets compared to a client-server approach.