total-agent-memory

概述

total-agent-memory是一个为AI编程助手设计的先进记忆系统，支持Claude Code、Codex CLI和Cursor等工具。它维护一个持久化的本地知识库，学习开发者的工作方式而非仅仅记录对话内容。系统利用时序知识图谱、程序记忆和基于AST的代码库摄取，提供跨会话和项目的上下文记忆。通过结合BM25、密集向量、图关系等多种检索技术，在LongMemEval基准测试中达到了96.2%的R@5准确率。

试试问 AI

装完之后，这里有 3 个你可以让 AI 做的事：

• memory_save_fast

  Save memory content quickly without LLM calls in fast mode

  试试:Remember that we chose pgvector over ChromaDB for multi-tenant RLS试试:Save this decision about using JWT-only session tokens
• memory_search_fast

  Search memory quickly with hybrid retrieval in fast mode

  试试:Why did we pick pgvector again?试试:Find information about vector database choices
• memory_explain_search

  Explain how a search query was processed

  试试:Explain how you found that memory about database choices试试:Show me the search pipeline for this query
• memory_warmup

  Warm up the memory system for faster initial access

  试试:Warm up the memory system试试:Preload memory caches
• memory_perf_report

  Get performance metrics for the memory system

  试试:Show me memory performance metrics试试:Get latency stats
• memory_rebuild_fts

  Rebuild the full-text search index

  试试:Rebuild the full-text search index试试:Update search indexes
• memory_rebuild_embeddings

  Rebuild all embedding indexes with current models

  试试:Rebuild embedding indexes试试:Update vector embeddings
• memory_eval_locomo

  Run LoCoMo benchmark on memory system

  试试:Run LoCoMo evaluation试试:Test memory accuracy with LoCoMo
• memory_eval_recall

  Evaluate memory recall performance

  试试:Evaluate memory recall试试:Test how well memory retrieves information
• memory_eval_temporal

  Evaluate temporal reasoning capabilities of memory

  试试:Evaluate temporal memory试试:Test memory timeline accuracy
• memory_eval_entity_consistency

  Evaluate consistency in entity extraction across sessions

  试试:Test entity consistency试试:Check how well entities are tracked over time
• memory_eval_contradictions

  Check for contradictions in memory knowledge

  试试:Find contradictions in memory试试:Check for inconsistent information

可对比工具

安装

pip install total-agent-memory

{
  "mcpServers": {
    "total-agent-memory": {
      "command": "python",
      "args": ["-m", "total_agent_memory.server"],
      "env": {}
    }
  }
}

total-agent-memory init
total-agent-memory serve

total-agent-memory 对比