RAG Frameworks · Dataojitori

nocturne_memory

Nocturne Memory is a Python-based MCP Server that provides LLM-agnostic long-term memory storage using graph-structured data, replacing vector RAG approaches. It persists AI personality, identity, and context across multiple models and sessions via SQLite or PostgreSQL backends with a visual dashboard for management.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/Dataojitori/nocturne_memory

1.3k

GitHub stars

154

Forks

Python

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	Dataojitori/nocturne_memory
Owner	Dataojitori
Primary language	Python
License	MIT — OSI-approved
Stars	1.3k
Forks	154
Open issues	4
Latest release	2.5.4 (2026-05-31)
Last updated	2026-06-26
Source	https://github.com/Dataojitori/nocturne_memory

What nocturne_memory is

MCP protocol server (Python 3.10+) implementing persistent memory via read_memory, search_memory, and write_memory tools backed by structured databases. Supports namespace isolation, rollback, and audit trails; stateless design enables AI migration across Claude, Gemini, GPT, and local models without memory loss.

Quickstart

Get the nocturne_memory source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/Dataojitori/nocturne_memory.git2cd nocturne_memory3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-Model AI Agents with Persistent Identity

Deploy the same AI personality across Claude, Gemini, and local models—each session retrieves consistent identity, preferences, and history from the central memory server. Ideal for teams switching models or building model-agnostic agent systems.

Long-Session Development Assistants

Enable Cursor, GitHub Copilot, or Claude Desktop to maintain project context, code patterns, and user preferences across weeks of development work without context window overflow or token exhaustion.

Autonomous AI Systems with Accountability

Build supervised agents that self-document decisions, reasoning chains, and state transitions in a human-auditable format. Dashboard review/rollback ensures human oversight before memory mutations take effect.

Implementation considerations

MCP client compatibility mandatory: verify your IDE/AI platform (Cursor, Claude Desktop, Antigravity) supports MCP stdio or SSE protocol before committing.
Database choice impacts scale: SQLite sufficient for single-user/dev; PostgreSQL required for production multi-agent systems or >10K memory nodes.
System Prompt critical: default MCP tool descriptions enable basic function discovery but don't coach AI to proactively use memory. Supply custom System Prompt (linked in docs) for autonomous memory management.
Frontend build adds ~30-60s first startup time (Node.js compilation); subsequent launches instant. Plan for initial delay in CI/CD.
Namespace isolation requires explicit configuration; default single namespace—design early if multi-personality or multi-tenant isolation needed.

When to avoid it — and what to weigh

Proprietary Single-Model Ecosystems — If your org is fully committed to ChatGPT or Claude's native memory/knowledge features and has no plans to switch models, the added complexity may not justify the decoupling benefit.
Unstructured Vector-First Workflows — If your use case is pure semantic search over unstructured documents (e.g., legal corpus discovery), traditional vector RAG or Pinecone will be simpler. Nocturne optimizes for graph-structured, relational memories.
Security-Isolated Multi-Tenant SaaS — Nocturne stores all AI memories in a single server. If you need cryptographically isolated memory per tenant with zero cross-contamination risk, a partitioned per-tenant solution is safer.
Minimal DevOps Tolerance — Requires Python 3.10+, Node.js, MCP client integration, and database management. If your team avoids local infra, managed SaaS solutions (with known trade-offs) may be preferable.

License & commercial use

MIT License. Permits commercial use, modification, distribution, and private deployment without attribution requirement (though attribution appreciated). No license incompatibilities with common OSI stacks.

MIT is a permissive OSI license that explicitly allows commercial use, closed-source derivatives, and proprietary deployment. You may run this in production, sell SaaS wrapping it, or embed it in commercial products without license restriction. No patent grants or liability clauses beyond standard MIT scope—review your own liability insurance for production use.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Strong
License clarity	Clear
Deployment complexity	Moderate
DEV.co fit	Good
Assessment confidence	High

Security considerations

No encryption-at-rest enforced; memories stored in plaintext in SQLite/PostgreSQL. API token authentication present but not mandatory (check config defaults). No audit log tamper-proofing; human review dashboard is primary control. MCP runs as subprocess with client process privileges—no privilege isolation. Recommended: TLS for networked Dashboard, VPN/firewall access control, encrypt DB filesystem, treat as trusted-network-only service. Do not expose Dashboard or MCP endpoint to untrusted networks without additional auth proxy.

Alternatives to consider

Anthropic Notebooks (Claude API)

Native integration with Claude, persistent across sessions, no infrastructure required. Trade-off: Claude-only, opaque storage, no self-audit visibility, limited graph structure.

LangChain Memory Agents + Vector DB (Pinecone/Weaviate)

Lighter-weight, multi-model capable via abstraction layer, mature ecosystem. Trade-off: vector-centric (not graph), no rollback/audit UI, requires prompt engineering for proactive recall.

Local LLM + SQLite (custom build)

Maximum control, no third-party dependency. Trade-off: no UI, high engineering cost, no MCP standardization, steep DevOps burden.

Software development agency

Build on nocturne_memory with DEV.co software developers

Nocturne Memory is open-source and ready to deploy. Start with a 5-minute local setup, explore the visual dashboard, then decide if multi-model memory architecture fits your roadmap. MIT license means zero restrictions on commercial use.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

basic-memory tradememory-protocol memory-bank-mcp

Related on DEV.co

Explore the category and the services that help you build with it.

RAG Frameworks AI Application Development Custom Software Development API Development Talk to DEV.co

nocturne_memory FAQ

Can I use Nocturne Memory with multiple AI models simultaneously (Claude + Gemini in parallel)?

Yes, via namespace isolation. Each AI gets its own memory namespace; both read/write to the same server. Concurrent writes within a single namespace require serialization (human audit or external lock). Plan multi-agent concurrency early.

What happens if my MCP client crashes?

The MCP server subprocess may remain running. Restart your AI client to reconnect. No automatic cleanup; check your system process list. Some clients (Claude Desktop) handle auto-restart; others don't. Test your workflow.

Is Nocturne Memory suitable for production SaaS?

Functional yes, but requires hardening: TLS, API auth enforcement, database encryption, audit log protection, and multi-tenant namespace design. Not a turnkey SaaS product; treat as a building block. Early-stage stability risk.

How do I migrate memories if I switch MCP servers or databases?

Export from Dashboard or direct DB dump, then import into new instance. No built-in migration tool in docs. Database schema stability unknown across major versions—test upgrades in dev first.

Software development & web development with DEV.co

Adopting nocturne_memory is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate rag frameworks software in production.

Ready to Give Your AI a Persistent Brain?

Talk to DEV.co Schedule a Consultation