MCP提交

Ollama MCP Bridge WebUI

一个连接本地LLM与MCP服务器的TypeScript桥梁项目，提供Web界面使开源模型能使用类似Claude的工具能力，支持文件系统、网络搜索和复杂推理等功能。

人工智能聊天机器人开发者工具 #本地AI #工具集成 #Web界面 #多协议 .TypeScript

2.5分

2025-04-28 17:40:27

概述

安装

内容详情

替代品

What is Ollama-MCP Bridge?

The Ollama-MCP Bridge is a TypeScript application that connects locally running language models (through Ollama) with various Model Context Protocol (MCP) servers. It provides a web interface that allows you to interact with open-source AI models while giving them access to powerful tools like filesystem operations, web search, and reasoning capabilities - similar to what commercial AI assistants offer.

How to use the Bridge?

After installation, you can access the web interface where you can chat with your local AI model. The system will automatically determine when to use tools like file access or web search based on your questions. You don't need to manually select tools - the AI will handle this intelligently.

Use Cases

Ideal for: 1) Privacy-conscious users who want AI assistance without cloud services, 2) Developers needing local AI with tool access, 3) Researchers experimenting with open-source models, 4) Anyone wanting to extend local AI capabilities with web search and file access.

Key Features

Multi-MCP IntegrationConnect and use multiple MCP servers simultaneously for different capabilities

Automatic Tool DetectionThe system intelligently selects the right tool based on your queries without manual intervention

Web InterfaceClean, user-friendly interface with collapsible tool descriptions and conversation history

Comprehensive ToolsetIncludes filesystem access, web search (via Brave), and advanced reasoning capabilities

Local OperationRuns entirely on your hardware with optional internet connectivity only for web searches

Pros and Cons

Advantages

Privacy: Your data stays on your local machine

Flexibility: Combine different tools as needed

Cost-effective: No API fees for local model use

Customizable: Adjust configuration to your needs

Offline capable: Most features work without internet

Limitations

Requires decent hardware to run local models effectively

Web search requires a Brave API key (free tier available)

Local models may be less capable than commercial cloud options

Initial setup requires some technical knowledge

Getting Started

Automatic InstallationRun the automatic installation script for easiest setup

Manual SetupIf you prefer manual control, follow these steps:

ConfigurationEdit the bridge_config.json and .env files to customize your setup

Running the BridgeStart the bridge and access the web interface

Example Uses

Research AssistanceFind and summarize recent information on a topic

File ManagementWork with local files through natural language

Complex Problem SolvingBreak down multi-step problems

Frequently Asked Questions

What models are supported?Any model that runs in Ollama. The default is Qwen, but you can configure others in bridge_config.json.

Do I need programming knowledge to use this?Basic setup requires some technical skills, but the web interface is designed for non-technical users once running.

How do I get a Brave API key?Sign up at brave.com/search/api - there's a free tier available for personal use.

Can I add my own tools?Yes, you can develop custom MCP servers and add them to the configuration.

Where are files stored?All files are kept in the workspace directory (../workspace by default) on your local machine.

Additional Resources

Ollama DocumentationOfficial documentation for Ollama

Model Context ProtocolMCP GitHub repository

Brave Search APISign up for Brave Search API keys

Qwen Model InfoDetails about the Qwen language models

精选MCP服务推荐

Duckduckgo MCP Server

已认证

DuckDuckGo搜索MCP服务器，为Claude等LLM提供网页搜索和内容抓取服务

Firecrawl MCP Server是一个集成Firecrawl网页抓取能力的模型上下文协议服务器，提供丰富的网页抓取、搜索和内容提取功能。

Framelink Figma MCP Server是一个为AI编程工具（如Cursor）提供Figma设计数据访问的服务器，通过简化Figma API响应，帮助AI更准确地实现设计到代码的一键转换。

Exa MCP Server是一个为AI助手（如Claude）提供网络搜索功能的服务器，通过Exa AI搜索API实现实时、安全的网络信息获取。

TypeScript

1,426

5分

Edgeone Pages MCP Server

EdgeOne Pages MCP是一个通过MCP协议快速部署HTML内容到EdgeOne Pages并获取公开URL的服务

百度地图MCP Server是国内首个兼容MCP协议的地图服务，提供地理编码、路线规划等10个标准化API接口，支持Python和Typescript快速接入，赋能智能体实现地图相关功能。

MiniMax Model Context Protocol (MCP) 是一个官方服务器，支持与强大的文本转语音、视频/图像生成API交互，适用于多种客户端工具如Claude Desktop、Cursor等。

Context7 MCP是一个为AI编程助手提供实时、版本特定文档和代码示例的服务，通过Model Context Protocol直接集成到提示中，解决LLM使用过时信息的问题。

TypeScript

4,852

4.7分

Ollama-MCP 桥接 WebUI

一个用 TypeScript 实现的桥梁，将本地 LLM（通过 Ollama）连接到模型上下文协议（MCP）服务器，并提供了一个网络界面。该桥接工具可以让开源模型具备与 Claude 相同的能力，从而实现实现完全在您自己的硬件上运行的强大本地 AI 助手。

特性

多 MCP 集成：同时连接多个 MCP 服务器
工具识别：根据查询自动识别应使用的工具
网络界面：带有可折叠工具描述的干净用户界面
全面工具集：文件系统、网络搜索和推理功能

安装

自动安装

最简单的设置桥接工具的方法是使用包含的安装脚本：

./install.bat

该脚本将执行以下操作：

检查并按需安装 Node.js
检查并按需安装 Ollama
安装所有依赖项
创建工作区目录（../workspace）
设置初始配置
构建 TypeScript 项目
下载 Qwen 模型到 Ollama

运行脚本后，您只需要执行以下操作：

将 API 密钥添加到 .env 文件中（配置文件中的 $VARIABLE_NAME 变量将被替换为实际值）

手动设置

如果您更喜欢手动设置：

从 ollama.com/download 安装 Ollama
拉取 Qwen 模型：ollama pull qwen2.5-coder:7b-instruct-q4_K_M
安装依赖项：npm install
创建工作区目录：mkdir ../workspace
配置 API 密钥到 .env 文件
构建项目：npm run build

配置

桥接工具通过两个主要文件进行配置：

1. bridge_config.json

此文件定义 MCP 服务器、LLM 设置和系统提示。使用 $VARIABLE_NAME 语法引用环境变量。

示例：

{
  "mcpServers": {
    "filesystem": {
      "command": "node",
      "args": [
        "To/Your/Directory/Ollama-MCP-Bridge-WebUI/node_modules/@modelcontextprotocol/server-filesystem/dist/index.js",
        "To/Your/Directory/Ollama-MCP-Bridge-WebUI/../workspace"
      ],
      "allowedDirectory": "To/Your/Directory/Ollama-MCP-Bridge-WebUI/../workspace"
    },
    "brave-search": {
      "command": "node",
      "args": [
        "To/Your/Directory/Ollama-MCP-Bridge-WebUI/node_modules/@modelcontextprotocol/server-brave-search/dist/index.js"
      ],
      "env": {
        "BRAVE_API_KEY": "$BRAVE_API_KEY"
      }
    },
    "sequential-thinking": {
      "command": "node",
      "args": [
        "To/Your/Directory/Ollama-MCP-Bridge-WebUI/node_modules/@modelcontextprotocol/server-sequential-thinking/dist/index.js"
      ]
    }
  },
  "llm": {
    "model": "qwen2.5-coder:7b-instruct-q4_K_M",
    "baseUrl": "http://localhost:11434",
    "apiKey": "ollama",
    "temperature": 0.7,
    "maxTokens": 8000
  },
  "systemPrompt": "You are a helpful assistant."
}