Agentic memory systems reuse past experience to boost future performance, yet most existing designs curate memory at write time: after a task finishes, its trajectory is distilled into a fixed artifact such as a reflection, workflow, skill, or reasoning strategy, which is later retrieved by similarity. This forces the system to decide what to remember before the future query is known, irreversibly discarding information and producing a query‑independent summary that must serve many downstream tasks. Learning such a write‑time curator is also hard because the value of a storage decision often becomes apparent only when a relevant query arrives, possibly many tasks later, creating a long‑horizon credit‑assignment problem.
We instead retain raw trajectories and defer curation until read time, when the current task is known. Given the retrieved traces and the new task, a memory curator synthesizes a compact, task‑adaptive payload tailored to the immediate need. Because this payload is consumed on the same task, the curator can be trained directly from immediate task success, avoiding delayed utility signals and the need to artificially group related tasks.
Across ALFWorld, WebShop, and $$\tau^2$$‑bench, our Just‑in‑Time Memory (JitMem) consistently outperforms no‑memory agents as well as heuristic and learned write‑time memory methods, improving over the strongest baseline by 16.2, 16.3, and 3.9 absolute success‑rate points, respectively. Notably, even an untrained curator is already competitive with or surpasses these baselines, showing that task‑adaptive read‑time curation itself is a major source of the gain; training the curator further compounds the improvement.
Review: JitMem’s strategy of postponing memory curation to read time effectively tackles the long‑horizon credit assignment challenge of write‑time curation, enabling more efficient, task‑specific memory utilization and demonstrating a viable path to enhance LLM agents in multi‑task settings.