> For the complete documentation index, see [llms.txt](https://moofai.gitbook.io/moof/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://moofai.gitbook.io/moof/moof-memory-network.md).

# MOOF Memory Network

Persistent memory is essential for agents that operate over long durations or repeatedly over many sessions. The **MOOF Memory Network** is a built-in distributed memory store that all agents in a MOOF universe can utilize for storing and recalling information. It combines multiple approaches to support different types of memory:

* **Episodic Memory Streams:** Agents keep a chronological *event log* of experiences, including observations, actions, outcomes, and messages. 
* **Semantic Vector Memory:** This offers a vector database storing text embeddings, extending an agent’s context window for unstructured knowledge. 
* **Structured Memory**: In the MOOF Knowledge Graph, Memory Networks store facts structured as triplets or JSON — for example, finalized contract terms.
* **Memory Management and Pruning:** Memory Networks prevent excessive memory growth by summarizing and compressing outdated information, similar to how agents create high-level reflections from raw data. The Evolution Engine helps determine which memories are valuable for learning. 

In practical terms, the Memory Network operates behind the scenes. A user in the Playground might simply tick a box “Enable long-term memory” for an agent, and MOOF handles the rest. Or a template might come pre-configured to log certain interactions.