A field guide to the agent memory gold rush

What it does Awesome-Agent-Memory is a curated directory of research and tooling around memory for large language and multimodal models. It catalogs papers, benchmarks, open-source products, tutorials, and news, all organized by whether the memory is nonparametric (external stores, graphs, vectors), parametric (model weights), or inspired by cognitive science. The goal is to map how different approaches let agents retain context, retrieve it later, and reason across sessions.

The interesting bit Instead of dumping everything into one bucket, the list splits memory into granular categories—text, graph, multimodal understanding, multimodal generation—and even dedicates sections to agent evolution and continual learning. It also biases toward reproducibility: open-source projects with public code are bolded and ranked higher than closed-source entries. The maintainers also quietly slip in their own drop-in Mem0 alternative, TeleMem, complete with a cheeky import telemem as mem0 tagline.

Key highlights

• Covers both LLM and MLLM memory, from plain-text retrieval to multimodal generation.
• Benchmarks are split across text, multimodal, and simulation environments.
• Open-source products are ordered by GitHub stars and visually prioritized in bold.
• Includes adjacent topics like cognitive science, context engineering, and reinforcement learning.
• Tracks industry news and workshops alongside academic papers.

Caveats

• The list mixes curation with self-promotion: the maintainers embed their own TeleMem project as a bolded, drop-in replacement for a competing tool.
• Several news headlines carry 2026 datelines with no context, making the timeline look suspect.
• Beyond one-line blurbs, the list offers little synthesis or guidance on how to choose between approaches.

Verdict Researchers and engineers building stateful agents who need a quick map of the memory landscape should bookmark this. If you are looking for deep analysis or a neutral, conflict-of-interest-free survey, look elsewhere.