Needle

Needle is an MCP server that provides managed retrieval-augmented generation (RAG) capabilities to search and extract context from custom document collections. It connects LLM interfaces directly to the Needle document management platform, allowing users to query data embedded in PDFs, DOCX, XLSX, and web pages using natural language. Knowledge workers, software developers, and research analysts use this server to index technical manuals, internal wikis, spreadsheets, and online articles into organized collections. Once connected, client assistants can read, organize, and query these collections directly within conversation windows. The server supports both hosted remote endpoints and self-hosted environments via local Python scripts or Docker containers. By handling semantic search and retrieval externally, it avoids tedious manual text copying and enables LLM agents to deliver grounded, accurate answers backed by multi-format source files across extensive personal or enterprise archives.

Category: AI Memory & Context

Tags: documents, rag, retrieval, search

Visit Needle

How to install and configure Needle

1. Obtain an API key from your Needle account settings. 2. Configure your client. For Claude Desktop, edit claude_desktop_config.json (~/Library/Application Support/Claude/claude_desktop_config.json on macOS or %APPDATA%/Claude/claude_desktop_config.json on Windows) to add the remote server: json { "mcpServers": { "needle": { "command": "npx", "args": [ "mcp-remote", "https://mcp.needle.app/mcp", "--header", "Authorization:Bearer ${NEEDLE_API_KEY}" ], "env": { "NEEDLE_API_KEY": "<your-needle-api-key>" } } } } 3. Alternatively, for local installations, clone https://github.com/needle-ai/needle-mcp, install uv, and point the config command to uv running needle-mcp with your NEEDLE_API_KEY. 4. Restart Claude Desktop to start using the tools.

What you can do with Needle

• Creating custom document collections to categorize technical documentation, project specs, and team research files using natural language prompts. - Indexing web URLs and external documents into specified Needle collections directly through assistant conversations. - Executing semantic searches across complex PDF, DOCX, and XLSX files to answer specific analytical questions. - Listing and inspecting all available document collections to audit knowledge base coverage from an AI chat interface.

Key facts

• https://github.com/needle-ai/needle-mcp
• AI Memory & Context, Files, Documents & PDFs
• documents, rag, retrieval, search

Part of MCP Servers

How do I install Needle MCP?

You can connect using the recommended remote server by adding an npx mcp-remote command entry to your Claude Desktop or Cursor configuration file pointing to the Needle streamable HTTP endpoint. Alternatively, you can clone the repository and run it locally using uv or Docker.

What can Needle MCP do?

It allows language models to manage and query document collections stored in Needle. Capabilities include creating new collections, listing existing collections, adding documents or web links to collections, and performing semantic searches across indexed files such as PDFs, spreadsheets, and word documents.

Which MCP clients work with Needle?

Needle MCP is designed and tested to work with Claude Desktop and Cursor. Any Model Context Protocol client capable of handling remote HTTP or local standard I/O server processes can also integrate with it.

Do I need a Needle API key to use this server?

Yes, an active Needle API key is required. You can obtain your key from your Needle account settings and add it to the environment variables section of your MCP client configuration file.