PlatypusDB is an agent-native bitemporal event database developed for WunderOS. It operates as a purpose-built component inside the agentic environment created by Pentad Labs, addressing the need for data storage that sits directly within an agent's execution substrate rather than behind a separate authority boundary.
The system uses a Merkle WAL as its single source of truth. Multiple derived views are generated from this log, including graph, vector, versioned-tree, image, and analogy representations. Queries and rules are supported through crisp Datalog reads as well as fuzzy VSA resonance. PlatypusDB maintains two clocks to provide bitemporal semantics. The entire database can be replayed deterministically and is tested using a TigerBeetle-style VOPR for fuzzing. Implementation consists of a Gleam control plane based on OTP running over a Zig data plane, with the design emphasizing one write path, two clocks, and zero glue code.
It is positioned for use with AI agents inside WunderOS, where writes, derived inferences, tool actions, and replay checkpoints can share a single ordering and failure model. The approach combines exact and vague recall capabilities within the same in-process database.
In the Other AI space, PlatypusDB takes a focused approach. It focuses on providing a database substrate that lives inside AI agents with deterministic replay, fuzzy recall, and unified failure models. PlatypusDB is an open-source project aimed at developers. The project is open source (Open Source). The product ships for the command line, and it can be self-hosted.
Pentad Labs builds and maintains PlatypusDB, and the product first shipped in 2026. Among its 4 catalogued features are Bitemporal Events, Merkle WAL, and Datalog Queries.
Latest indexed changes and source events
Knowing, Remembering, Exactly, Vaguely: An Agent-Native Database (PlatypusDB) verified by the PulseGate indexer
Other apps tracked under the same category.
