Welcome to VecLabs
VecLabs makes Recall - the complete memory layer for AI agents. Two packages. @veclabs/solvec stores, encrypts, and verifies agent memory. @veclabs/recall retrieves and assembles structured context. (Phase 7)Quickstart
Get your first vector stored and queried in under 5 minutes.
What is a Vector Database?
New to vector databases? Start here.
Why VecLabs?
Understand what makes VecLabs different from Pinecone, Qdrant, and Weaviate.
Use Cases
See what you can build with VecLabs.
What makes VecLabs different
Speed
4.7ms p99 at 100K vectors, 1536 dimensions. Rust HNSW in-process - no
network round-trip, no garbage collector.
Privacy
Vectors are encrypted with AES-256-GCM before leaving the SDK. VecLabs
cannot read your data.
Verifiability
After every write, a cryptographic fingerprint is recorded. Anyone can
verify your data has not been tampered with.
Install
Get your API key at app.veclabs.xyz, then:Your first query in 30 seconds
Current status
| Component | Status |
|---|---|
| Rust HNSW core | Complete - 57 tests, 4.7ms p99 |
| Solana Anchor program | Live on devnet |
| WASM bridge | Complete |
| AES-256-GCM encryption | Complete |
| Shadow Drive | Complete |
| Memory Inspector | Complete - Phase 6 |
| Hosted API (api.veclabs.xyz) | Complete - Phase 8 |
| Developer Dashboard (app.veclabs.xyz) | Complete - Phase 8 |
| Context Retrieval (@veclabs/recall) | Planned - Phase 7 |
Phases 1–6 and Phase 8 are complete and stable.
@veclabs/recall (Phase 7) is in development.