Cognitive Behavioral Therapy (CBT) is a first‑line evidence‑based treatment for depression, yet clinicians spend considerable time on pre‑session preparation, post‑session documentation, and longitudinal tracking of cognitive pathology, limiting scalability. This work introduces a clinician‑facing AI decision‑support system that integrates a multi‑agent CBT framework (MACBT) with a CBT‑specific longitudinal memory module (CD Memory). MACBT encodes the five CBT stages—assessment, Socratic questioning, cognitive restructuring, behavioral experiments, and treatment monitoring—into five collaborative agents that share information. CD Memory logs the type, frequency, severity, and restructuring efficacy of cognitive distortions across sessions, producing pre‑session pathology reports and intervention‑priority recommendations. A Chinese CBT dialogue corpus was generated via dual‑role large language model simulation, used to fine‑tune a Qwen3‑14B backbone with supervised learning and direct preference optimization. GPT‑4 judges show MACBT surpasses MeChat, SoulChat, PsyChat, and CPsyCounX in professionalism (2.62) and clinical authenticity (2.25). Adding the memory module improves overall session quality by 12.6% and yields a mean cross‑session continuity, intervention progression, and personalization score of 2.29.
Review: By preserving the full CBT workflow while incorporating cross‑session memory, the system achieves higher clinical practicality and personalization, illustrating the promise of combining multi‑agent architectures with long‑term memory.