OpenWebUI MCP Server
A bridge between LLMs and OpenWebUI workspaces that enables managing projects, notes, and knowledge bases. It supports safe project configuration updates through a propose-evaluate-apply workflow and direct file uploads to knowledge bases.
README
OpenWebUI MCP Server
A Model Context Protocol (MCP) server for managing projects, notes, and knowledge bases in OpenWebUI.
Overview
open-webui-mcp provides a bridge between LLMs and your OpenWebUI workspace. It allows models to:
- List and manage projects.
- Create and update project-specific notes.
- Associate knowledge bases with projects.
- Propose and apply updates to project configurations based on chat history.
- Upload files directly to knowledge bases.
Prerequisites
- Bun (v1.0+)
- OpenWebUI instance
- OpenRouter API key
Quick Start
-
Clone and install:
git clone https://github.com/your-repo/open-webui-mcp.git cd open-webui-mcp bun install -
Configure environment variables:
export OPENWEBUI_BASE_URL="http://localhost:3000" export OPENWEBUI_API_KEY="your-api-key" export OPENROUTER_API_KEY="your-openrouter-key" -
Build and run:
bun run build bun run start
Project Mapping
The server uses a projects.json file to map project IDs to OpenWebUI folders and resources. This allows for a structured way to manage project-specific context.
Example projects.json
{
"version": "1.0",
"projects": [
{
"id": "project-1",
"name": "My Project",
"folder_id": "folder-uuid",
"note_ids": [],
"knowledge_base_ids": []
}
]
}
Safety Model (Propose/Evaluate/Apply)
To ensure safe updates to project configurations, the server implements a three-step workflow:
- Propose: The model analyzes a conversation and generates a
ProposedChangeSet. - Evaluate: An LLM (via OpenRouter) evaluates the proposal for safety and relevance, assigning a score (0-1).
- Apply: If the proposal is deemed safe (score > threshold), it can be applied to the project.
Documentation
Troubleshooting
- Connection Errors: Ensure
OPENWEBUI_BASE_URLis accessible and the API key is correct. - Permission Denied: Check if your OpenWebUI API key has sufficient permissions to manage folders, notes, and knowledge bases.
- Evaluation Failures: Ensure your
OPENROUTER_API_KEYis valid and has credits. - Missing Projects: Verify that
projects.jsonexists in yourDATA_DIRand is valid JSON.
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。