LanceDB

LanceDB builds data infrastructure for AI teams to store, search and prepare multimodal datasets for model training and applications. Its platform brings raw data, metadata and embeddings together, supporting dataset curation, feature engineering, and vector, full-text and hybrid search with SQL filters. LanceDB OSS is an Apache 2.0 licensed embedded retrieval library; LanceDB Enterprise provides a commercial multimodal lakehouse deployable in a private cloud or customers’ own cloud environments. Applications include retrieval-augmented generation, recommendation systems and agent memory.

Founded in 2022, LanceDB is led by co-founders Chang She, CEO, and Lei Xu, CTO. Its technical foundation is Lance, a columnar format supporting random access and shuffling for multimodal training data, including images and point clouds. LanceDB builds disk-based search indexes on that format; both are written in Rust. Apache Arrow interoperability and automatic versioning help teams work across data tools and track dataset changes. Lance introduced community governance in 2025, separately from the LanceDB project.

In June 2025, the company reported more than 20 million downloads of its open-source packages over the preceding year, with LanceDB Enterprise customers including Runway, Midjourney and Character.ai searching tens of billions of vectors and managing petabytes of training data. That year, it closed a $30 million Series A led by Theory Ventures. By August 2026, LanceDB also shipped as CrewAI’s default vector backend for agent memory.