Neotoma

Neotoma is a software product that provides deterministic, versioned state and durable memory for multi-agent systems. It takes the form of a state layer that allows agents to share and persist information across tools with provenance tracking, schema constraints, and audit logs. The product is aimed at multi-agent builders and operators who want to prevent context loss, inference variance, and silent data mutations across different sessions and interfaces.

• Durable memory that persists across sessions and tools
• Cross-tool synchronization for agents across Claude, Cursor, ChatGPT, and MCP-connected tools
• Versioned history that preserves original records and prevents retries from silently overwriting data
• Auditable change logs and tamper detection that hold over months and years
• Schema constraints that reject malformed records to stop errors from compounding silently
• Reproducible reconstruction to return corrupted databases to a known-good state
• Access through a CLI, REST API, Inspector app, or MCP-connected tools

Pros

• Provides cross-tool continuity so operators do not have to act as a human sync layer between different AI tools
• Maintains state integrity with versioned history and provenance tracking instead of relying on standard chat memory or retrieval-augmented generation
• Enables replay and debugging by allowing developers to inspect, diff, and replay agent memory
• Prevents silent data mutations and schema errors by enforcing structured history and constraints

Cons

• Pricing is not published on the page and requires contacting the creators
• The product is explicitly not built for one-off thought-partner workflows or note-taking apps
• Open source status is not recorded in the available facts
• Requires developers to use or integrate specific interfaces like CLI, REST API, or MCP-connected tools

Neotoma fits developers, pipeline operators, and agentic app builders running multi-agent systems who need reliable state management across multiple sessions and tools. It is suited for teams dealing with pipeline state, cross-tool sync, and debugging complex agent interactions. It is a poor fit for one-off thought-partner workflows, hobbyist chatbots, or simple note-taking applications.

Pricing is not published on the page, and the listing states that you have to contact them for pricing.

Unlike standard chat memory, Retrieval-Augmented Generation, or accumulated Markdown and JSON files that suffer from drift and silent conflicts, Neotoma provides deterministic state with versioning, tamper detection, and provenance. Rather than forcing developers to build and maintain custom MCP servers or local SQLite databases, it ships pre-built guarantees for agent state integrity. It operates as a unified state layer across multiple disparate agent interfaces.

Claude, ChatGPT, Cursor, OpenCode, OpenClaw, IronClaw, Claude Code, REST API, CLI, and MCP-connected tools.

Neotoma is worth trying for developers and teams running serious production multi-agent systems who suffer from context loss and inference variance across tools. It provides robust guarantees like versioned history, schema constraints, and cross-tool sync that replace custom-built memory systems. However, because pricing requires contacting the company and the tool is explicitly overkill for hobbyist chatbots or note-taking apps, casual users should skip it.