remio-harness-plugin
MCP server that exposes remio knowledge base tools—search, read, RAG, web retrieval, note and collection management, sharing, and agent execution—to MCP-compatible clients. It invokes the local remio CLI directly with safety confirmations for state-changing operations.
README
remio Harness Plugin
MCP server for using a local remio knowledge base from DeepSeek Harness and any MCP-compatible agent client.
It exposes remio search, reading, RAG, web retrieval, note management, collection management, sync-folder management, public sharing, agent execution, and advanced syscalls. It invokes the local remio executable directly with argument arrays; it never opens a shell.
Requirements
- Node.js 20 or later
- pnpm (required by
dsh plugin) - remio desktop app installed and running
- remio CLI available on
PATH(or setREMIO_CLI_PATH)
Install
After the package is published, install it in a DeepSeek Harness profile:
dsh plugin --profile standard add @remio/harness-plugin
DSH loads the bundle automatically and exposes the remio MCP tools under the mcp__remio__ namespace, for example mcp__remio__remio_search_notes and mcp__remio__remio_ask.
The same published package can also be used directly by any MCP-compatible client:
{
"mcpServers": {
"remio": {
"command": "npx",
"args": ["-y", "@remio/harness-plugin"]
}
}
}
For local development:
npm install
npm run build
node dist/index.js
{
"mcpServers": {
"remio": {
"command": "node",
"args": ["/absolute/path/to/remio-harness-plugin/dist/index.js"]
}
}
}
Tools
| Tool | Purpose | Confirmation |
|---|---|---|
remio_help |
Local CLI usage | No |
remio_search_notes |
Search the knowledge base | No |
remio_read |
Read a note or supported local file | No |
remio_ask |
RAG answer with citations | No |
remio_web |
Commercial web search or clean web extraction | No |
remio_notes |
Create, update, delete, attach images, status, permissions | Required |
remio_collections |
Add or remove a note from a collection | Required |
remio_sync_folders |
List or manage indexed folders | Required for changes |
remio_share |
Publish, refresh, or revoke a public note link | Required |
remio_run_agent |
Run a remio Agent task | Required |
remio_syscall |
Advanced remio syscall escape hatch | Required |
remio_cli |
Full CLI compatibility entry point | Required for changes |
The structured tools are the recommended interface. remio_cli is retained for a remio CLI capability that has not yet been promoted into a dedicated structured tool.
Safety model
The server uses MCP tool annotations so compatible clients can show appropriate risk prompts. It additionally refuses every state-changing call unless the caller sends confirm: true after obtaining explicit user approval.
The server does not bypass remio authorization. Existing remio restrictions, passwords, session-scoped write checks, and public-sharing behavior remain enforced by the local remio app.
Configuration
| Variable | Meaning | Default |
|---|---|---|
REMIO_CLI_PATH |
Absolute path to the remio executable | remio |
REMIO_CLI_TIMEOUT_MS |
CLI invocation timeout | 120000 |
Development and release
npm ci
npm run check
npm test
npm run pack:check
The CI workflow verifies Node.js 20, 22, and 24. Before the first public release, add the canonical public repository URL to package.json, publish under the @remio npm organization, and submit the npm package plus repository link to the applicable Harness plugin directory. DSH detects the package as a native profile bundle through its dsh.bundle.patch manifest field.
The DSH bundle has been verified against @deepseek-ai/dsh@0.1.0-rc.6: a clean profile installed the package successfully and its composed configuration included the remio-mcp bridge row.
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。