eCBRa

The agent that brings your world into problem/solution pairs and finds the best solution according to your preferences.

eCBRa stands for easy Case-Based Reasoning agent. We call it e-Zebra! It is living in the Cyber Savannah—a digital landscape where neon lightning meets the quiet rhythm of the savannah.

Release Date: 2026 April 9
Latest version: 1.0
Editors:
Andreas Korger, University of Würzburg, DE
Contact:
(Collaborations and contributions welcome:)
info[at]piri-safety.com

Abstract

eCBRa provides a high-performance agentic Case-Based Reasoning (CBR)^* framework designed for transparency and precision. Unlike "black-box" models, eCBRa is built for scenarios requiring full explainability of agent behavior, allowing users to trace every decision back to specific, stored cases. The framework is highly maintainable and context-aware, enabling setups that are finely tuned to individual parameters and evolving environmental data. eCBRa offers maximum deployment flexibility: it can function as a standalone intelligent agent or be exported to integrate seamlessly with third-party ecosystems and modern agentic paradigms, such as Model Context Protocol (MCP), OpenClaw, and other orchestrators.

^*Why it works

Think of Case-Based Reasoning (CBR) as "reasoning by analogy." Instead of relying on a set of rigid "if-then" rules or complex statistical weights, a CBR system solves new problems by remembering how it solved similar ones in the past. It follows a fundamental human logic: "Similar problems have similar solutions." The 4-Step "CBR Cycle" To understand how it works, developers usually refer to the 4R Cycle: Retrieve: When a new problem occurs, the system searches its memory (the "Case Base") for the most similar past cases. Reuse: It takes the solution from those past cases and applies it to the new problem. Revise: If the old solution doesn't fit perfectly, the system adapts or "tweaks" it to better suit the current context. Retain: Once the problem is solved successfully, the new experience is stored as a new "case," making the agent smarter for next time. Why eCBRa is different from LLMs While a Large Language Model (like an LLM) "guesses" the next word based on patterns it learned during training, a CBR agent (like your eCBRa) looks at a specific database of facts. Explainability: If someone asks "Why did the agent do that?", eCBRa can point to a specific file and say, "Because in Case #402, which had 90% similar parameters, this action worked." Maintenance: You don't need to "retrain" the whole brain to update it. You just add, delete, or edit a single case (a JSON or YAML file) in the library. In short, it’s the difference between a student who has memorized a textbook (LLM) and a veteran professional who has a filing cabinet full of past project reports (CBR).