什么是 Memos管理技能？

Memos技能旨在弥合AI智能体与个人知识管理系统之间的差距。它提供了一个强大的接口，用于与自托管的Memos实例（usememos/memos）进行交互，允许用户将他们的AI视为数字图书管理员。通过将其纳入Openclaw Skills库，开发人员可以将灵感捕获、技术片段和会议摘要直接自动化地存入其私有基础设施中。

该技能利用openclaw-memos-mcp服务器处理智能体与Memos API之间的安全通信。它确保每条笔记、查询请求和更新都能得到精确处理，使Openclaw Skills成为重视数据主权和有序个人信息的用户的核心组件。

下载入口:https://github.com/openclaw/skills/tree/main/skills/sinhong2011/usememos

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

Memos管理技能 应用场景

• 在开发过程中捕获稍纵即逝的想法或代码片段。
• 使用高级CEL过滤表达式检索特定的项目笔记。
• 自动化创建私密每日日志或任务日志。
• 修改笔记可见性或置顶重要信息以便更好地整理。
• 通过托管删除工作流清理过时的文档。

1. AI智能体接收到保存、查找或编辑笔记的自然语言请求。
2. 智能体从Openclaw Skills工具包中选择合适的工具，如memos_create或memos_list。
3. 该技能使用提供的API URL和访问令牌与openclaw-memos-mcp服务器通信。
4. 服务器向Memos实例执行REST API调用，以执行请求的CRUD操作。
5. 智能体接收结构化响应，向用户确认操作，并处理任何必要的后续步骤。

Memos管理技能 配置指南

要在您的环境中激活此技能，必须配置MCP服务器。将以下代码块添加到您的配置文件中：

{
  "mcpServers": {
    "memos": {
      "command": "npx",
      "args": ["openclaw-memos-mcp"],
      "env": {
        "MEMOS_API_URL": "http://your-memos-instance:5230",
        "MEMOS_TOKEN": "your_access_token_here"
      }
    }
  }
}

保存后，重启您的客户端以启用这些Openclaw Skills功能。请确保您的令牌是通过Memos UI中的“设置 > 访问令牌”生成的。

Memos管理技能 数据架构与分类体系

该技能使用以下输入和输出模式与Memos数据模型交互：

name: memos
description: >
  Create, read, update, and delete memos on a Memos instance (usememos/memos).
  Handles requests like "save this as a memo", "list my recent memos",
  "update memo #123", or "delete memo #456".
  Uses openclaw-memos-mcp for all operations.

Memos

What it does

Provides full CRUD operations on a self-hosted Memos instance through 5 MCP tools. Agents can create quick notes, search existing memos, update content or visibility, and delete memos they no longer need.

Inputs needed

• For creating: content (required), visibility (optional, defaults to PRIVATE)
• For listing: filter expression (optional), page size (optional)
• For get/update/delete: memo ID (required)

Prerequisites

openclaw-memos-mcp MCP server

This skill requires the openclaw-memos-mcp MCP server to be running. Before using any memos_* tool, check if the tools are available. If not, tell the user they need to add the MCP server to their configuration:

{
  "mcpServers": {
    "memos": {
      "command": "npx",
      "args": ["openclaw-memos-mcp"],
      "env": {
        "MEMOS_API_URL": "http://localhost:5230",
        "MEMOS_TOKEN": ""
      }
    }
  }
}

Tell the user to:

1. Replace MEMOS_API_URL with their Memos instance URL
2. Get an access token from Memos: Settings > Access Tokens > Create
3. Replace with the token
4. Restart their MCP client after saving the configuration

Workflow

Creating a memo

Call memos_create with the content and optional visibility.

• content: Markdown text. Supports #hashtags which Memos auto-extracts as tags.
• visibility: PRIVATE (default), PROTECTED (logged-in users), or PUBLIC (everyone).

Listing memos

Call memos_list to browse or search memos.

• pageSize: Number of results (default 20).
• pageToken: For pagination, use the nextPageToken from a previous response.
• filter: CEL filter expression. Examples:
  • tag == "work" — memos tagged #work
  • visibility == "PUBLIC" — public memos only
  • creator == "users/1" — memos by a specific user

Reading a specific memo

Call memos_get with the memo ID. The ID is the part after memos/ in the resource name (e.g., if the name is memos/abc123, the ID is abc123).

Updating a memo

Call memos_update with the memo ID and the fields to change. Only specify fields you want to update — unspecified fields remain unchanged.

• content: New Markdown content
• visibility: New visibility level
• pinned: true to pin, false to unpin

Deleting a memo

Call memos_delete with the memo ID. This is irreversible. Always confirm with the user before deleting.

Guardrails

• Default visibility to PRIVATE — never create public memos unless explicitly asked
• Validate content is non-empty before creating
• Confirm with the user before deleting any memo
• On 401/403 errors, tell the user to check their MEMOS_TOKEN
• On connection errors, tell the user to check their MEMOS_API_URL
• When listing returns empty results, suggest the user check their filter or confirm the Memos instance has data