Memory Bank FAQs

Name: Memory Bank
Author: Protocol-Lattice

Question 1

Can multiple AI agents share memory using Memory Bank?

Accepted Answer

Yes, Memory Bank features 'Shared Memory Spaces' where multiple agents or users can collaborate. These spaces come with fine-grained access control (ACLs) and time-to-live (TTL) support for grants and spaces.

Question 2

What embedding models are compatible with Memory Bank?

Accepted Answer

Memory Bank uses `AutoEmbedder` from the Go Agent Framework, allowing dynamic configuration of embedding models via environment variables. This includes popular models like OpenAI and Gemini, as well as various local models.

Question 3

Which database backends does Memory Bank support for memory storage?

Accepted Answer

Memory Bank supports multiple vector store backends, including PostgreSQL (with pgvector), Qdrant, MongoDB (with Atlas Vector Search), and an In-Memory option for quick tests and ephemeral storage.

Question 4

What is Memory Bank?

Accepted Answer

Memory Bank is a production-ready Model Context Protocol (MCP) server offering a powerful, vector-native memory bank for AI agents. It provides persistent, searchable, and shareable memory, built with the Protocol-Lattice Go Agent Framework.

Question 5

How does Memory Bank manage different types of AI agent memory?

Accepted Answer

It provides rich memory management, including a short-term buffer for temporary session-based recall, long-term persistence for embedded memories, and contextual retrieval to fetch relevant memories from both sources based on a query.

Memory Bank

Memory Bank

주요 기능

사용 사례

주요 기능

사용 사례