vectorize-mcp-server

MCP.Pizza Chef: vectorize-io

Three tools do the work: search a collection of documents you set up beforehand on Vectorize's website, run a longer research pass over that collection which can optionally pull in the live web, and pull the readable text out of a file. Everything happens on Vectorize's servers, so you need an account there plus three separate settings — an organisation ID, a token and a collection ID — and the server will not start unless all three are present.

Data

Files/PDF

Web/Research

Use This MCP server To

Ask a question across every document I have uploaded Get a longer written answer with its sources listed Pull the readable text out of a PDF Check what my own files say before answering a client Research a topic using my documents and the live web

README

Vectorize MCP Server

A Model Context Protocol (MCP) server implementation that integrates with Vectorize for advanced Vector retrieval and text extraction.

Installation

Running with npx

export VECTORIZE_ORG_ID=YOUR_ORG_ID
export VECTORIZE_TOKEN=YOUR_TOKEN
export VECTORIZE_PIPELINE_ID=YOUR_PIPELINE_ID

npx -y @vectorize-io/vectorize-mcp-server@latest

VS Code Installation

For one-click installation, click one of the install buttons below:

vectorize-mcp-server FAQ

Do I need an account?

Yes. It only works against a document collection you have already built on Vectorize's website, and it needs three separate settings from there before it will start.

Does it cost money?

The server itself is free, but Vectorize is a commercial service and the searching, research and extraction all run on their platform under your plan there.

Can I use this to search my company's documents?

Yes, provided those documents are already loaded into a Vectorize collection. It searches that collection, not the files sitting on your computer.

Do my documents leave my machine?

Yes. Searching, research and text extraction all happen on Vectorize's servers, so treat confidential material accordingly.

Which apps does it work in?

VS Code has a one-click install button, and Cursor, Windsurf, Cline and Claude Desktop all take the same short config block.

Can it get stuck?

Yes. Research and extraction wait for a result in a loop with no time limit, so a job that never finishes leaves the request hanging indefinitely.

Can I control how documents are broken up?

No. Text extraction uses a fixed size and a fixed placeholder file name, neither of which can be changed from the conversation.

Is it maintained?

Yes. Vectorize publish it themselves, it is MIT licensed, and it was last updated in June 2026.