Agent memory is the capacity of an AI agent to store, recall, and reuse information across interactions, typically divided into short-term memory (the working context of a single session) and long-term memory (persistent storage that survives across sessions). Long-term memory is usually held outside the model in databases or vector stores and retrieved when relevant.
Why it matters
Without persistent memory, an AI agent forgets everything once its context window clears, forcing the user to re-establish background every time. Long-term memory lets an agent build on past research, remember a client's situation, and stay consistent over weeks or months, which is foundational for production agents doing ongoing knowledge work. It is the difference between a chatbot that resets and an assistant that accumulates understanding.
How Pith relates
Pith functions as a durable, external long-term memory grounded in what you have actually read and saved, rather than in conversation history an assistant might lose. Because it is queryable over MCP, an AI agent can treat your cited wiki and bookmarks as persistent recall that outlives any single chat. The memory is source-backed, so what the agent remembers is traceable to where it came from.
See also
Last reviewed: 7 June 2026 · Licensed CC BY 4.0 · cite freely with attribution to Pith.