Website

What it does

Neotoma is a state management layer designed for AI agents and applications that require persistent, auditable memory. It stores versioned entities such as contacts, tasks, transactions, and decisions with complete provenance tracking and immutable observation logs. The tool enables schema-first data extraction and maintains a comprehensive history of all state changes.

How to set up

Neotoma functions as an MCP (Model Context Protocol) server that can be installed and configured within Claude, Cursor, ChatGPT, and other compatible AI platforms. Installation involves: 1) Cloning the repository from GitHub, 2) Installing dependencies, 3) Configuring the local SQLite database, 4) Adding the MCP server configuration to your AI platform, and 5) Authenticating cross-client access for multi-platform support.

Key features

• Versioned entity storage with full audit trails
• Immutable observation logs and complete provenance tracking
• Local-first SQLite database for privacy and performance
• Cross-client memory synchronization across Claude, Cursor, ChatGPT, and OpenAI
• Schema-first data extraction and validation
• Deterministic state management for reliable agent behavior