Wolbarg

Wolbarg is a TypeScript software development kit designed to provide shared semantic memory for AI agents. It addresses the challenge of fragmented or transient memory in multi-agent systems by enabling multiple agents to write to and search from a single, persistent memory store. This shared memory is local-first, supporting both SQLite and PostgreSQL as backends, and is accessed through a compact TypeScript API. The SDK is intended for scenarios where agents require durable, concurrent memory with the ability to recall information based on meaning rather than just exact keys.

A distinguishing feature of Wolbarg is its use of semantic search via embeddings, allowing agents to retrieve information using natural language queries. Unlike setups that rely on in-process memory, JSON files, or generic key-value stores, Wolbarg enables persistent storage and concurrent access without the need for external services like Redis or custom locking mechanisms. Its API is centered around simple commands such as remember() and recall(), streamlining the process of storing and retrieving information.

Wolbarg is model-agnostic and can be installed via npm, making it accessible to developers working in TypeScript environments. It supports both local and shared deployment models, depending on whether SQLite or PostgreSQL is used. The SDK does not require a separate memory SaaS, emphasizing local-first operation and crash safety. It is suitable for developers building agent-based systems that need reliable, meaningful memory recall across multiple processes or machines.

evidence_sufficient=true