Memori - Structured memory from agent trace & execution

Memori Labs launched its new agent-native memory infrastructure, enabling agents to create structured, long-term memory directly from the agent trace — including execution paths, tool results, workflow steps, outcomes, and decision-making logic. This allows memory to be generated not only from what an agent says, but from what an agent actually does.

Unlike conversational memory wrappers or vector retrieval layers, Memori structures memory from both conversation and agent execution — turning tool calls, decisions, and workflow traces into persistent, queryable state. Memori's benchmark results reflect the approach: 81.95% accuracy on LoCoMo using only 1,294 tokens per query, roughly 5% of full-context cost, saving users up to 95% on inference costs.

The open-source project has grown to more than 15K GitHub stars and 200K downloads.

Bessemer Venture Partners has identified memory and context management as a key part of the emerging AI infrastructure harness layer, citing Memori as one of the category leaders.

Видео Memori - Structured memory from agent trace & execution канала Memori Labs