Context Lens

Name: Context Lens
Author: cornelcroi

@cornelcroi

关于 Context Lens

Semantic search knowledge base for MCP-enabled AI assistants. Index local files or GitHub repos, query with natural language. Built on LanceDB vector storage. Works with Claude Desktop, Cursor, and other MCP clients.

#semantic #lancedb #search #knowledge-base

基本信息

分类

其他

许可证

MIT

运行时

python

传输方式

stdio

发布者

cornelcroi

提交者

Corneliu

配置

使用下面的配置,将此服务器添加到你的 MCP 客户端。

{
  "mcpServers": {
    "context-lens": {
      "command": "uvx",
      "args": [
        "context-lens"
      ]
    }
  }
}

工具

未检测到工具

工具是从 README 中自动提取的。维护者可以在 ## Tools 标题下列出工具,即可填充这部分内容。

概览

What is Context Lens?

Context Lens is a self-contained MCP server with built-in serverless vector storage (LanceDB) that transforms any content—codebases, documentation, contracts, or text files—into a searchable knowledge base for AI assistants. It enables semantic understanding over keyword matching, allowing the AI to answer questions about contextually relevant information even when the exact terms are absent.

How to use Context Lens?

Add Context Lens to your MCP client configuration (Kiro IDE, Cursor, Claude Desktop, Continue.dev) using uvx context-lens. For programmatic use, invoke the server with StdioServerParameters(command="uvx", args=["context-lens"]) and interact via the Python SDK. After setup, add content by asking the AI (e.g., "Add all Python files from ./src") and query naturally (e.g., "How does the authentication system work?").

Key features of Context Lens

Semantic search that understands meaning, not just keywords
Zero setup: no installation, configuration, or API keys required
Serverless vector storage with LanceDB (file-based, no cloud)
100% local and private: all data stays on your machine
Index local files or public GitHub repositories
Smart parsing: language-aware chunking for code and documents

Use cases of Context Lens

Query a codebase: ask how authentication works, find database models, or compare frameworks without manual browsing
Learn from open source: index a GitHub repository and ask in-depth technical questions about its internals
Analyze documents: scan contracts for termination clauses, payment terms, or specific obligations
Fetch files from URLs: instantly add a remote file (e.g., a README) and ask questions about its content

FAQ from Context Lens

What makes Context Lens different from keyword search tools?

Context Lens uses semantic embeddings (384-dimensional vectors) to capture meaning. It finds "authentication"-related code even if the word never appears, unlike traditional grep or keyword search.

What are the runtime requirements and dependencies?

Python 3.11+ and the uvx runner. The embedding model (~90MB) downloads on first use. All processing is local with no external API calls.

Where does my data live?

Entirely on your machine—stored as a LanceDB file on disk. No data is sent to the cloud; the server is fully offline.

What are the known limitations?

Maximum file size per document is 10 MB (configurable via MAX_FILE_SIZE_MB). Only text-based file types are supported (25+ extensions). The system works via stdio transport; no HTTP or custom authentication.