Vectra MCP

An MCP server based on TypeScript for interacting with the Vectra knowledge base, providing functions such as creating collections, embedding texts/files, and querying the knowledge base.

Knowledge management and memory Search tools #Knowledge base management #Text embedding #Collection query #File processing .TypeScript

rating : 2 points

downloads : 7

update time : 2025-04-28

What is the Vectra MCP Server?

The Vectra MCP Server is a middleware service that allows applications to interact with the Vectra knowledge base through a standardized protocol. It provides a set of tools for creating collections, embedding texts and files, performing intelligent searches, and other operations.

How to use the Vectra MCP Server?

By sending requests conforming to the MCP protocol to the server, you can perform various knowledge base operations. First, you need to install and start the server, and then use the provided toolset for interaction.

Use cases

Suitable for scenarios that require managing large - scale document knowledge bases, implementing intelligent search functions, or building question - answering systems, especially when combining vector search and keyword search.

Main features

Collection managementCreate and manage different knowledge base collections to organize different types of documents and content.

Text embeddingConvert text content into vector representations and store them in the knowledge base, supporting batch processing.

File processingRead local file content directly and embed it into the knowledge base, supporting the addition of metadata.

Hybrid searchCombine the advantages of vector search and keyword search to provide more accurate search results.

Graph search enhancementEnhance search results using graph database relationships, supporting the setting of search depth and relationship types.

Advantages and limitations

Advantages

Provide a standardized MCP protocol interface for easy integration

Support multiple content embedding methods, including text and files

The hybrid search algorithm combines the advantages of vector and keyword search

Graph search enhancement provides more intelligent result associations

Limitations

Require a separately running Vectra backend service

May require an additional encapsulation layer for non - technical users

Large - scale file processing may require performance optimization

How to use

Installation and startup

Install the Node.js environment and start the MCP server

Create a collection

Create a new knowledge base collection to organize your content

Add content

Add content to the collection by embedding text or files

Query the knowledge base

Use the hybrid search function to query relevant content in the collection

Usage examples

Build a product documentation search systemEmbed product documentation into the knowledge base, and users can find relevant documentation content through natural language queries

Research paper knowledge baseEstablish a collection of academic papers, and researchers can search for research in relevant fields

Frequently Asked Questions

What kind of backend support does the MCP server need?

How to handle the upload and embedding of large files?

How are the search results sorted?

Related resources

MCP protocol specification

The official specification document of the Model Context Protocol

Vectra API documentation

Detailed interface description of the backend Vectra API

Example code repository

Example code and integration cases for using the Vectra MCP Server

🚀 Vectra MCP Server

A server based on the Model Context Protocol (MCP) for interacting with the Vectra knowledge base.

This MCP server, written in TypeScript, provides tools for managing and querying Vectra instances and allows integration with MCP-compatible clients. It interacts with the backend Vectra API (assumed to be running separately).

✨ Features

Tools

This server exposes the following tools for interacting with Vectra:

create_collection: Create a new Vectra collection.
- Input: name (string, required), description (string, optional)
list_collections: List existing Vectra collections.
- Input: None
embed_texts: Embed multiple text items in batches into Vectra.
- Input: items (an array of objects containing text (required) and optional metadata), collectionId (string, optional)
embed_files: Read multiple local files and embed their contents into Vectra.
- Input: sources (an array of local file paths, required), collectionId (string, optional), metadata (object, optional - applied to all items)
add_file_to_collection: Add an embedded file (referenced by its ID) to a specific Vectra collection.
- Input: collectionId (string, required), fileId (string, required)
list_files_in_collection: List the files in a specific Vectra collection.
- Input: collectionId (string, required)
query_collection: Query the knowledge base of a specific Vectra collection.
- Note: This tool always uses hybrid search (vector + keyword) and enables graph search enhancement by default.
- Input: collectionId (string, required), queryText (string, required), limit (number, optional), maxDistance (number, optional), graphDepth (number, optional), graphRelationshipTypes (array of strings, optional), includeMetadataFilters (array of objects, optional), excludeMetadataFilters (array of objects, optional)
delete_file: Delete a file and its associated embedded vectors.
- Input: fileId (string, required)
get_arangodb_node: Retrieve a specific node directly from the underlying ArangoDB database by its key.
- Input: nodeKey (string, required - e.g., chunk_xyz or doc_abc)