Agent memory
Storage an agent reads and writes across steps or sessions, outside the context window.
Memory lets an agent persist facts, preferences and progress beyond a single context. Short-term memory is the scratchpad within a run; long-term memory is a store (vector database, key-value, graph, plain files) queried before each step. Benchmarks such as LoCoMo and LongMemEval measure retrieval quality over long conversations. Stack Overflow found Redis (42.9%) and GitHub's MCP server (42.8%) are the most-used memory/data tools among agent builders.
Agent memory in our posts
- AI assistant that can control your computer: five that do it, and what each one can reach
- AI assistant that remembers everything: what ChatGPT, Claude, Hermes and TODO for AI actually keep
- Best AI personal assistants in 2026: which ones answer, and which ones act
- 94.7% on LoCoMo — and why most of the gap between published memory numbers isn't the memory