remindb

概述

Remindb 作为 MCP 服务器运行，将非结构化笔记转换为结构化的索引记忆系统。它能解析多种格式（Markdown、HTML、JSON、YAML、TOON）为类型化节点树，每个节点都有温度跟踪、令牌优化和版本控制功能。系统提供一系列 MCP 工具，包括用于浏览的 MemoryTree、用于快速 FTS5 搜索的 MemorySearch、用于增量更新的 MemoryDelta，以及用于知识图谱导航的 MemoryRelated。

试试问 AI

装完之后，这里有 5 个你可以让 AI 做的事：

• MemoryTree

  Get a tree view of the memory structure with typed nodes and temperatures

  试试:Show me the memory tree structure试试:What's the hot content in my notes right now?
• MemorySearch

  Search the memory using FTS5 with ranked results and token budgeting

  试试:Find all information about payment processing试试:Search for documentation on the data model
• MemorySummarize

  Summarize cold nodes to reduce token usage while preserving structure

  试试:Summarize the cold parts of my project documentation试试:Compact this stale information to save context
• MemoryDelta

  Get changes since last cursor for efficient resync

  试试:What's new in my notes since last time?试试:Show me the updates to the documentation
• MemoryRelated

  Find knowledge graph relations between nodes up to 5 hops

  试试:What's related to the architecture section?试试:Show me content connected to payment integration
• MemoryRelate

  Create new knowledge graph relationships between nodes

  试试:Connect the data model section to the schema documentation试试:Link this observation to the relevant user feedback

可对比工具

安装

curl -fsSL https://raw.githubusercontent.com/radimsem/remindb/main/install.sh | bash

iwr -useb https://raw.githubusercontent.com/radimsem/remindb/main/install.ps1 | iex

git clone https://github.com/radimsem/remindb.git
cd remindb
go build -o ~/.local/bin/remindb ./cmd/remindb

{
  "mcpServers": {
    "remindb": {
      "command": "remindb",
      "args": ["serve"]
    }
  }
}

FAQ

remindb 如何减少令牌使用？

通过将笔记解析为具有基于哈希的重复数据删除和基于温度的相关性的结构树，允许代理仅访问内存的相关部分，而不是重新读取整个文件。

remindb 支持哪些文件格式？

支持 Markdown、HTML、JSON、YAML 和 TOON 格式，并在有益的情况下自动压缩。

remindb 对比