context-bridge
Provides LLMs with secure, read-only access to local documentation by scanning directories, extracting content from PDF, DOCX, Markdown, and text files, and performing keyword searches.
README
<p align="center"> <img src="./assets/logo.png" width="200" height="200" style="border-radius: 50%;" alt="Context Bridge Logo"> </p>
<p align="center"> <a href="https://www.npmjs.com/package/@elkraps/context-bridge"> <img src="https://img.shields.io/npm/v/%40elkraps%2Fcontext-bridge?color=blue" alt="NPM Version"> </a> </p>
Context Bridge MCP
Context Bridge is a Model Context Protocol (MCP) server designed to provide Large Language Models (LLMs) with secure, read-only access to local documentation. It acts as an intermediary layer, allowing agents to scan directories, search for keywords, and extract content from PDF, DOCX, Markdown, and plain text files directly from the host filesystem.
This tool resolves the context isolation problem by enabling agents to reference large local knowledge bases without requiring file uploads or manual copy-pasting.
Features
- File System Scanning: recursively list documents with metadata (size, modification date) to understand the knowledge base structure.
- Content Extraction: parse and extract text from binary formats (PDF, DOCX) and text-based formats (Markdown, TXT).
- Semantic Search: perform keyword-based search across multiple files to locate relevant information snippets.
- Safety: operates in read-only mode to prevent accidental data modification.
Available Tools
-
list_documents: Scans a directory for supported files (PDF, DOCX, MD, TXT).
path(string, optional): Absolute path to the directory. Defaults to current.recursive(boolean, optional): Enable subdirectory scanning.
-
read_document: Extracts text content from a file. Handles binary conversion automatically.
path(string, required): Absolute path to the file.
-
search_documents: Performs case-insensitive keyword search with context snippets.
query(string, required): Search term.path(string, optional): Directory scope.recursive(boolean, optional): Enable subdirectory search.
Getting Started
Local Installation (Source)
-
Clone the repository:
git clone https://github.com/elkraps/context-bridge-mcp.git cd context-bridge -
Install dependencies and build:
npm install npm run build -
Client Configuration: Add the server configuration to your MCP-compatible client's settings file. For example:
{ "mcpServers": { "context-bridge": { "command": "node", "args": ["/ABSOLUTE/PATH/TO/context-bridge/build/index.js"] } } }
NPX Usage
Configure your MCP client to run the server directly via npx:
{
"mcpServers": {
"context-bridge": {
"command": "npx",
"args": ["-y", "@elkraps/context-bridge"]
}
}
}
Usage
To effectively utilize this tool, you must explicitly direct the agent to interface with the local documentation using the context-bridge terminology.
System Prompt Configuration
To ensure the agent prioritizes this MCP server over generic shell commands (like ls or cat), add the following instruction to your System Prompt, .cursorrules, or custom instructions:
"You have access to local documentation via the
context-bridgetools. ALWAYS uselist_documents,read_document, andsearch_documentsto explore, read, or search files. Do not use shell commands for documentation tasks."
Interaction Examples
Correct Prompting:
"Please analyze the system architecture described in my documentation folder at
/Users/username/projects/docsusing context-bridge. List the available files first."
"Search for 'authentication protocols' within the local documentation using context-bridge and summarize the findings."
"Read the file
/Users/username/projects/docs/api-spec.pdfusing context-bridge and generate a Python client based on it."
Supported File Types
- PDF (
.pdf): Text extraction only (OCR not supported). - Microsoft Word (
.docx): Text extraction. - Markdown (
.md): Native text reading. - Plain Text (
.txt): Native text reading.
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。