turbopuffer
Serverless vector and full-text search on object storage
turbopuffer is profiled here as a Vector Database tool for engineering teams. Read about features, pricing, and how it compares to related options in the tools directory.
Description
turbopuffer is a serverless search database founded in 2023 by Simon Eskildsen and Carl Sverre. It places its primary index on object storage with SSD and memory caching above it, which lowers the cost of holding very large vector and text datasets while keeping warm queries fast. The architecture suits multi-tenant products with millions of namespaces, and companies including Cursor and Notion run search on it. Companies including Cursor and Notion run production search on it, which reflects its fit for very large multi-tenant workloads. Pricing tied to storage and queries lets products with many small namespaces keep costs predictable.
Key Capabilities:
Object-storage-native indexing with tiered SSD and memory caching
Vector similarity search with metadata filtering
BM25 full-text search and hybrid retrieval
Namespace-per-tenant design for large multi-tenant workloads
High write throughput with strong consistency on reads
Usage-based pricing tied to storage and queries