HelixDB

HelixDB is a database built for AI memory. It combines a property graph with native vector search and full text search in a single Rust engine running on object storage, so an application can traverse relationships and then run semantic or keyword search over the records it reaches, all in one request, instead of stitching a graph store and a vector store together. The argument is that agent systems need both shapes of query at once. A support question about whether an account has hit an error before traverses from the account to its past tickets, then runs vector and BM25 search over only those tickets. The same two step pattern covers long term memory of users, conversations and decisions so people do not have to repeat themselves, a connected company brain linking code, documents, people and projects so an answer arrives with its owner and runbook attached, and retrieval for AI features grounded in the records a given user is permitted to see. The site walks through that pattern across devtools, support, legal, finance, healthcare, sales, security, insurance, education, recruiting and media, and shows the database feeding models from several providers. The engine is open source and the company is backed by Y Combinator. Pricing is stated briefly rather than on a pricing page. Self hosting or embedding the database is free, and Helix Cloud plans start at $5 per month covering 200,000 reads and 100,000 writes. Larger volumes and enterprise terms are handled through the enterprise contact link in the navigation. One detail worth noting: the home page offers a setup prompt to copy into a coding assistant, so the intended path to a first working instance runs through an agent rather than a manual install.