Short-Term Memory

Session-scoped memory that persists within a conversation but resets between sessions, implemented through conversation history, scratchpad patterns, and working state that agents maintain during multi-step task execution. Short-term memory is the simplest form of agent memory and is built into every chat-based agent by default — it's literally the messages array passed to each API call. The critical architectural challenge is managing short-term memory as conversations grow long: naive approaches include the entire conversation history, which eventually exhausts the context window, while sophisticated systems use summarization, sliding windows, or relevance-weighted pruning to keep the most important information accessible. Understanding short-term memory matters because it's the foundation on which all other memory types build, and getting it wrong means your agent loses track of what it was doing mid-task. This concept connects to context window budget for the constraint that limits short-term memory capacity, memory management for strategies to keep memory within bounds, and long-term memory for what happens when information needs to survive across sessions.