Recall is a vector database built for AI agents that need memory they can prove. Most vector databases store your embeddings on their servers. You trust their uptime, their pricing, and their word that nothing has changed. Recall is built differently. Every write does three things:Documentation Index
Fetch the complete documentation index at: https://docs.veclabs.xyz/llms.txt
Use this file to discover all available pages before exploring further.
- Encrypts your vectors with AES-256-GCM using a key derived from your Solana wallet — before they leave your machine
- Persists the encrypted vectors permanently to Arweave via Irys (Pro and above)
- Anchors a SHA-256 Merkle root of all vector IDs to a Solana Anchor program — a cryptographic fingerprint of your collection, immutable and publicly verifiable
Architecture
8xjQ2XrdhR4JkGAdTEB7i34DBkbrLRkcgchKjN1Vn5nP
Key properties
- Client-side encryption — keys are derived from your Solana wallet. VecLabs cannot read your data.
- Tamper-evident — the on-chain Merkle root proves what vectors exist and in what state.
- Fast — 4.7ms P99 at 100K vectors on the Rust HNSW core.
- Cheap — 25/mo Pro with permanent storage.