In-Memoria

概述

In Memoria 是一个用 Rust 构建的 MCP 服务器，它能从实际代码库中学习并跨会话记忆。它分析模式、架构和约定，为 AI 编码工具提供上下文感知的辅助。该系统结合了对 12 种语言的 Tree-sitter AST 解析与语义分析和模式学习，将数据存储在本地 SQLite 和 SurrealDB 中。它为 AI 助手提供 13 种专业工具，包括项目蓝图、智能文件路由和模式预测功能。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• analyze_codebase

  Analyze files/directories with concepts, patterns, complexity

  试试:Analyze the main components in this project试试:What patterns exist in the utils directory?
• search_codebase

  Multi-mode search (semantic/text/pattern)

  试试:Find authentication-related code试试:Search for database connection patterns
• learn_codebase_intelligence

  Deep learning to extract patterns and architecture

  试试:Learn from my current codebase structure试试:Extract all patterns from this project
• get_project_blueprint

  Instant project context with tech stack and entry points

  试试:What's the architecture of this project?试试:Show me the tech stack and main entry points
• get_semantic_insights

  Query learned concepts and relationships

  试试:How is authentication related to user management?试试:What components handle API routing?
• get_pattern_recommendations

  Get patterns with related files for consistency

  试试:What patterns should I follow for error handling?试试:Show me consistent patterns for API endpoints
• predict_coding_approach

  Implementation guidance with file routing

  试试:How would I implement a password reset feature?试试:Where should I add a new caching layer?
• get_developer_profile

  Access coding style and work context

  试试:What's my coding style preference?试试:What was I working on last session?
• contribute_insights

  Record architectural decisions

  试试:Record that we're using JWT for authentication试试:Note our decision to use TypeScript for new features
• auto_learn_if_needed

  Smart auto-learning with staleness detection

  试试:Check if codebase needs updating试试:Quickly analyze any new changes
• get_system_status

  Health check

  试试:Is the memory system working properly?试试:Check system health status
• get_intelligence_metrics

  Analytics on learned patterns

  试试:Show me what patterns have been learned试试:How much of the codebase has been analyzed?

可对比工具

安装

# 全局安装
npm install -g in-memoria

# 或直接使用 npx
npx in-memoria --help

{
  "mcpServers": {
    "in-memoria": {
      "command": "npx",
      "args": ["in-memoria", "server"]
    }
  }
}

FAQ

In Memoria 支持哪些编程语言？

目前支持 12 种语言：TypeScript、JavaScript、Python、PHP、Rust、Go、Java、C/C++、C#、Svelte 和 SQL。

它如何存储数据？

使用 SQLite 存储结构化数据模式和元数据，使用 SurrealDB 及 SurrealKV 后端存储持久化向量嵌入。所有数据都保存在本地机器上。

Hacker News 讨论

开发者社区最近的相关讨论。

• Show HN: Built a memory layer that stops AI agents from forgetting everything▲ 13
  帖子 by pi_22by7 · 2025-08-19
• Show HN: In Memoria – AI memory layer for persistent context in coding tools▲ 2
  帖子 by pi_22by7 · 2025-08-31
• Show HN: In Memoria – MCP server that stops AI assistants from forgetting▲ 2💬 1
  帖子 by pi_22by7 · 2025-11-04

In-Memoria 对比