Vectra MCP Server

Example Use Cases

• 1

  Ingest and index text and local files into a Vectra knowledge base for later semantic retrieval.

• 2

  Perform hybrid vector+keyword searches with optional graph-based expansion across a specific Vectra collection.

• 3

  Programmatically manage collections and files (create/list collections, add files to collections, list files, delete embeddings) and inspect raw ArangoDB nodes for debugging.

Description

This TypeScript-based MCP server exposes tools to create and manage Vectra collections, embed texts and files in batch, perform hybrid (vector + keyword) queries with graph-search enhancements, and administer stored files and underlying ArangoDB nodes. It supports local file ingestion, batch text embedding, adding embedded files to collections, listing and deleting files, and direct retrieval of ArangoDB nodes for debugging. Designed to run alongside a separate Vectra API backend, it enables agents and applications to integrate retrieval and knowledge-management capabilities via the MCP protocol.