SapixDB

SapixDB is an agent-native database built around data domains owned by AI agents. It is described as a database for AI agents rather than humans, and it is framed as addressing data silos, weak structural trust, opaque AI lineage, the lack of an agent-native layer, and schema rigidity.

Its core design centers on three stated pillars: living AI agents as tables, cryptographic hash-linked strands, and governed self-evolution. The page says every data domain is owned by an AI agent with its own identity, Ed25519 keypair, and semantic understanding of its data. Every record is cryptographically signed and hash-linked, with lineage, tamper-evidence, and provenance treated as structural properties of the storage format. Schema changes are meant to emerge through governed agent negotiation, with mutant meta-agents observing production, proposing changes, testing them in an isolated sandbox, and delivering signed reports to human administrators for approval.

SapixDB also lists several concrete capabilities. It offers realtime Server-Sent Events streams for subscribing to agent writes, with callbacks in Python, TypeScript, Go, and the browser. It supports time travel queries for every state of every record, HIPAA-related PHI field classification with encrypted storage and access audit logs, SOX dual-sign enforcement for schema mutations, live agent health scores, schema drift detection, and exportable cryptographic proofs for records that can be verified without database access using Ed25519 public keys. The interface described in the page includes a control plane and SDK use for time travel, and it mentions queries expressed as goals through SaQL.

The page presents it as a database product and a “living database” for AI systems. It is offered with a free trial, and the navigation also includes a pricing section, docs, and a start free trial link.