cff-version: 1.2.0
message: "If you use TMCRA, please cite this software. OpenAI Codex assisted with development and reproducibility engineering."
title: "TMCRA — Agent Memory Engine"
type: software
version: "2026.07.18"
date-released: 2026-07-18
license: Apache-2.0
authors:
  - family-names: Yu
    given-names: Haoxin
    alias: reshuibuduo
    website: "https://github.com/reshuibuduo"
  - name: "OpenAI Codex"
    website: "https://openai.com/codex/"
repository-code: "https://github.com/reshuibuduo/TMCRA-Agent-Memory"
url: "https://github.com/reshuibuduo/TMCRA-Agent-Memory"
abstract: >-
  TMCRA is an agent memory engine that builds scope-isolated, source-traceable
  memory from long-running conversations and compiles relevant evidence for
  downstream agents. This release includes the algorithm snapshot, learned
  graph-scoring artifacts, and a reproducible LongMemEval pipeline.
keywords:
  - agent memory
  - long-term memory
  - memory graph
  - retrieval
  - LongMemEval