AgentIndex icon
AgentIndex
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Submit Tool
ToolsMCP ServersHOTCategoriesTrendingNewBest ListsCompareGuides
Home/
Memory & Context/
nocturne_memory
nocturne_memory logo

nocturne_memory

Active·★ 1.4k·MIT·Updated 2026-09-22

Nocturne Memory is an MCP server that gives an AI agent long-term memory across sessions, models and tools. Instead of a vector database it uses a deliberately simple design: SQLite by default (PostgreSQL for remote or multi-device use), with two core tables, memories and paths, so one memory can be reached through several URI paths such as `core://agent/my_user`.

A priority weight makes the agent re-read core memories at startup, special entries such as `system://boot`, `system://index` and `system://recent` load identity, the full index and recent changes, and the read, create and search tools understand the path hierarchy. Every change the AI makes takes a snapshot first, and a React web interface lets a human browse memories and roll back in one click.

It runs as three parts: a FastAPI backend, the MCP server (stdio or SSE) and the web UI. Setup is clone, `pip install -r backend/requirements.txt`, copy `.env.example` and set an absolute database path; the README even provides a prompt you can give your assistant to install it. The README is written in a dramatic style ("memories are for the soul") and is mainly in Chinese, with an English version. MIT licensed; our data shows the last push on 2026-09-22.

This write-up is based on the project's README and public repository data. We have not hand-tested the tool. Data as of 2026-10-16.

License
MIT
Latest release
2.5.6 · 2026-08-09
Last commit
2026-09-22
Contributors
11
Forks
179
Open issues and PRs
5
First published
2025-12-25

Source: GitHub, npm and PyPI public APIs.

#AI Memory Management#LLM Persistence#Model Context Protocol (MCP)#Structured Memory#SQLite#PostgreSQL#Identity Anchoring#Version Control
$ Install
$ git clone https://github.com/Dataojitori/nocturne_memory.git && cd nocturne_memory && pip install -r backend/requirements.txt
↗ Visit site★ GitHub
01

Features

01Long-Term Persistence: AI memories are no longer limited by token count or session boundaries, stored permanently on disk.
02Identity Anchoring: A priority-weighted system ensures AI consistently "recalls" core memories upon startup, reinforcing its identity.
03Associative Recall: Memories are interconnected via URI paths and aliases, forming a human-like associative network for flexible recall.
04Version Control: Automatic snapshots for every memory modification allow human owners to review changes and perform one-click rollbacks.
05Content-Path Separation: A unique architecture separates memory content from its access paths, enabling flexible aliases, versioning, and secure deletion.
02

Why choose it

+Simple storage (SQLite or PostgreSQL) with URI paths and aliases instead of a vector database.
+Automatic snapshots before every AI edit, with one-click rollback in a web UI.
+Priority weights and boot entries reload core memories at startup.
+MCP over stdio or SSE; MIT licensed.
03

Trade-offs

!No semantic or vector search, so recall depends on how well memories are organised into paths.
!Manual setup from source, with an absolute database path to configure and a separate wrapper for some clients.
!The README's tone is more manifesto than documentation, and it is mainly in Chinese.
04

Maintenance check

ActiveChecked: 2026-10-18
Last commit
2026-09-22
Commits, last 90 days
20
Releases, last 90 days
1
Latest release
2.5.6 (2026-08-09)
Open issues
5
Recent issues with at least one comment
25/30
Commits per week, last 12 weeks
Commits per week, last 12 weeks: 5, 1, 1, 6, 0, 0, 0, 2, 0, 0, 0, 0

Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.

Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.

Most discussed open issues

  • #32 缺少 MCP Resource 导致额外的 read_mcp_resource 调用10 comments · 2026-03-25
  • #42 审查发现：删除节点后 glossary 触发词未级联清理，导致悬挂引用累积3 comments · 2026-05-11
  • #55 挺好用的，后续还会更新吗2 comments · 2026-07-17
  • #61 豆啊，nocturne它拔屌无情，不想理你了lol 。ai真冷血1 comments · 2026-08-30
  • #25 佬，可以提供CLI命令的入口吗？这样就不需要mcp了1 comments · 2026-03-18

Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.

05

Compatibility

Python
Runtime
Verified via docs
SQLite
Database
Verified via docs
PostgreSQL
Database
Verified via docs
Model Context Protocol (MCP)
Protocol
Verified via docs
06

Quick start

1
$ git clone https://github.com/Dataojitori/nocturne_memory.git
2
$ cd nocturne_memory
3
$ pip install -r backend/requirements.txt
07

Use cases

↳Maintaining an AI agent's consistent long-term identity and persona across multiple interactions.
↳Personalizing user interactions by allowing AI to remember past conversations, preferences, and shared history.
↳Managing specialized knowledge bases for creative projects, such as character psychology for a novel or game mechanics.
↳Persistent storage for AI's acquired knowledge, insights, and evolving understanding across different sessions.
↳Enabling human owners to audit, review, and roll back AI's memory modifications through a visual dashboard.
08

Alternatives

⚡Top VS:
VSclaude-mem-lite★67→VSlocus★15→VSOri-Mnemos★331→VSmatryca-plumber★100→
claude-mem-lite logo
claude-mem-lite★ 67
Persistent long-term memory for Claude Code via MCP — captures coding decisions, bugfixes, and context across sessions. Hybrid FTS5 + TF-IDF search with episode batching. Single SQLite DB, no external services. Alternative to claude-mem with 600x lower cost.
VSCompare→
locus logo
locus★ 15
Persistent project-aware memory for AI coding tools. Built on MCP. Works with Codex, Claude Code, and other MCP-compatible clients.
VSCompare→
Ori-Mnemos logo
Ori-Mnemos★ 331
Local-first persistent agentic memory powered by Recursive Memory Harness (RMH). Open source must win.
VSCompare→
matryca-plumber logo
matryca-plumber★ 100
Bringing Andrej Karpathy's LLM Wiki to the Outliner Paradigm. Turn any AI Agent into a spatial Knowledge Architect using Logseq's atomic nodes.
VSCompare→
origin logo
origin★ 86
Local-first Rust daemon for AI work. Git-versioned memories, distilled wiki pages, sessions for Claude Code, Cursor, Codex, and any MCP client.
VSCompare→
home-memory logo
home-memory★ 77
MCP server that lets your AI assistant remember everything about your home.
VSCompare→
llm-wiki-kit logo
llm-wiki-kit★ 61
Stop re-explaining your research to your AI agent. Persistent, LLM-maintained wikis that compound over time. Drop PDFs, URLs, YouTube — your agent remembers forever. Based on Karpathy's LLM Wiki pattern.
VSCompare→
hippo-memory logo
hippo-memory★ 784
Biologically-inspired memory for AI agents. Decay, retrieval strengthening, consolidation. Zero runtime deps, SQLite, MCP. Benchmarked retrieval with an opt-in hosted TypeSafe Jev reranker.
VSCompare→
See all alternatives →

Comments

Log in to leave a comment

No comments yet. Be the first!

On this page
01Features02Why choose it03Trade-offs04Maintenance check05Compatibility06Quick start07Use cases08Alternatives
Dataojitori /
nocturne_memory
↗

MIT MCP memory server that stores agent memories as URI-addressed records in SQLite or PostgreSQL, with snapshots and a web UI for rollback

1.4k179MIT
Last commit1mo ago
StatusActive
CategoryMemory & Context
Contributors11
Latest release2.5.6
Links
Documentation↗Discussion↗Issues↗Releases↗
Cloud Infrastructure
Ad

Deploy on DigitalOcean

Host your MCP servers, Docker containers, and AI agents with $200 free credit over 60 days.

Claim $200 Free Credit
© 2026 AgentIndex.app|Built by a 10-year iOS Developer.
Best by use caseAboutContactAdvertisePrivacy PolicyTerms of UseQYSGitHubBuy me a coffee ☕

Browse by Category

MCP ServersCode AssistantWorkflow AutomationRAG / Knowledge BaseMulti-AgentBrowser AutomationLLM InfraDev ToolingObservability

Not affiliated with Anthropic, OpenAI or Microsoft.