Stuart C. Shapiro and colleagues at the State University of New York at Buffalo developed and maintained the knowledge representation, reasoning, and acting (KRRA) system known as SNePS.

The mental entities that some agent imagines and holds as true are the intensional domain of mental entities that are the focus of this model. Since any two terms that differ syntactically may have marginally different Fregean senses, the lack of an inherent equality operator serves as the primary means.

SNePS has three inference styles: 

• Formula-based, 
• Slot-based, and 
• Path-based 

The above styles are from their logic-based, frame-based, and network-based personalities. However, all three are interdependent and work together.

It is possible to use SNePS as a standalone KRR system. Following the GLAIR agent architecture, it has also implemented the mind of intelligent agents (cognitive robots) and its integrated acting component (a layered cognitive architecture). The agent designation used by the SNePS Research Group is Cassie.

Logic-based SNePS

SNePS KB is a logic-based system with terms, functions, and formulas. The set of logical connectives and quantifiers used by first-order logic takes one or more arbitrarily-sized arguments. Propositions are first-class entities in the intended domain, so formulas are proposition-denoting functional terms. SNePSLOG, the logic-based face of SNePS's input-output language, looks like naive logic because function symbols (including "predicates") and formulas (actually proposition-denoting terms) can be function arguments and quantified over. SNePS is a first-order logic that reifies function symbols and formulas. In addition, formula-based inference uses a natural-deduction-style inference engine with rules for connectives and quantifiers. SNePS formula-based inference is sound but lacks rules for natural language understanding and commonsense reasoning.

Furthermore, a SNePS KB proposition-denoting term may or may not be "asserted." SNePS is a paraconsistent version of relevance logic, so contradictions imply nothing. SNeBR, the SNePS Belief Revision subsystem, will detect any explicit contradiction and prompt the user to fix it. SNeBR removes the assertion status of any proposition.

Frame-based SNePS

A slotted frame represents every SNePS functional term (including proposition-valued terms). SNePSUL is an input-output language for SNePS's frame-based system, and it has three modes:

1. The system determines each functional term's case frame (set of slots) in two modes.
2. In mode 3, the user declares each function's case frame.
3. Any proposition-valued structure implies the frame with any of its slots filled by a subset of its fillers.

Unfortunately, this isn't always sound in the current implementation.

Network-based SNePS

The original meaning of "SNePS" was "The Semantic Network Processing System" because it is a propositional semantic network and a network-based system. Every node denotes a mental entity, some of which are propositions. According to the intended interpretation, every proposition in the network is by the node representing it. Some nodes in the SNePS logic are variables that only apply to other nodes.

Although a path-based inference rule states that if a given path exists from node n to node m, then some labelled arc r may be inferred as present from node n to node m. There is a sizable collection of recursive path constructors available.

Applications

SNePS has a wide range of KRR tasks, including natural language comprehension and generation, commonsense reasoning, and cognitive robotics. It's in several KR courses around the world. SNePS is a platform-independent system written in Common Lisp that is freely available.