生产级内存操作系统，专为代理型人工智能设计。通过MCP、兼容OpenAI网关和原生/v1/* REST与所有主要框架互操作——包括Claude Code、OpenClaw、ZeroClaw、Hermes,...
概览
该项目的权威源代码始终存在于 GitLab：该 GitHub 仓库保留用于托管 mnemos-enterprise 容器镜像 ghcr.io。开发、问题和合并请求均在 GitLab 上进行。 它不仅仅是放置字节的地方：它是一个命名子系统的运行时，管理代理内存在整个生命周期内，跨服务提供者、代理和时间范围——写入、嵌入、搜索、压缩、版本、推理、审计、联合、导出、导入和操作。 - 一个带有CLI优先部署表面的方案——六后端持久化层——默认为SQLite + sqlite-vec，PostgreSQL + pgvector，Oracle AI Database 26ai，IBM Db2 12.1，MySQL 9.0 Enterprise/HeatWave，以及MariaDB 11.7+。每个后端在首次连接时自行配置其模式。参见持久性。
概述（AI 翻译）
📍 正规来源：GitLab
本项目的权威来源始终存在于 GitLab 上 — https://gitlab.com/ncz-os/mnemos
这个 GitHub 仓库仅保留用于托管
mnemos-enterprise容器镜像 ghcr.io。开发、问题和合并请求均在 GitLab 上进行。
记忆 + 格雷亚
MNEMOS v7.0.5 是用于严肃代理操作的内存操作系统。 它是 它不仅仅是放置字节的地方：它是管理的命名子系统的运行时 代理内存在服务提供者、代理和时间跨度的全生命周期—— 写入、嵌入、搜索、压缩、版本、推理、审计、联合、导出， 导入，并操作。
盒子里有什么：
- 一个打包的FastAPI运行时，带有CLI优先的部署表面
- EPIMONE，六个后端持久化层 — SQLite + sqlite-vec 默认、PostgreSQL + pgvector、Oracle AI 数据库 26ai、IBM Db2 12.1、MySQL 9.0 Enterprise/HeatWave，以及MariaDB 11.7+。每个后端都能自我配置 首次连接时的模式。参见持久性。
- GRAEAE 推理总线和 PANTHEON 统一的 LLM 外观
- 操作员审计的压缩堆栈
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
📍 Canonical source: GitLab
The authoritative source for this project lives on GitLab — always: https://gitlab.com/ncz-os/mnemos
This GitHub repository is retained only to host the
mnemos-enterprisecontainer image on ghcr.io. Development, issues, and merge requests happen on GitLab.
MNEMOS + GRAEAE
MNEMOS v7.0.5 is the memory operating system for serious agentic work. It is not just a place to put bytes: it is a runtime of named subsystems that manage the full lifecycle of agent memory across providers, agents, and time horizons — write, embed, search, compress, version, reason over, audit, federate, export, import, and operate.
What is in the box:
- a packaged FastAPI runtime with a CLI-first deployment surface
- EPIMONE, the six-backend persistence layer — SQLite + sqlite-vec by default, PostgreSQL + pgvector, Oracle AI Database 26ai, IBM Db2 12.1, MySQL 9.0 Enterprise/HeatWave, and MariaDB 11.7+. Every backend self-provisions its schema on first connect. See Persistence.
- the GRAEAE reasoning bus and PANTHEON unified LLM facade
- an operator-audited compression stack
- a divergent dream-state pipeline: REPLAY → CLUSTER → CONSOLIDATE → SYNTHESISE → EXTRACT
- a GDPR right-to-be-forgotten worker
- the PERSEPHONE archival subsystem
- first-party CHARON portability: MIF/MPF import and export, universal ingest, migrate-in adapters, and STYX encrypted off-fleet backups
- KRONOS recall observability
How it is packaged. MNEMOS ships as a small core (
mnemos-core) plus separately installablemnemos.*namespace subsystems (GRAEAE, PANTHEON, KNEMON) and the standalone STIPHOS hive service. CHARON’s portability, migration, ingestion, and STYX backup code became first-party core modules on 2026-09-18; only Docling’s heavy document-conversion dependencies remain optional throughmnemos-core[docling]. The published container image isghcr.io/ncz-os/mnemos-enterprise— a single multi-arch (amd64 + arm64) manifest with every backend driver (Oracle, MySQL, MariaDB) except Db2, which is amd64-only. Pin an exact version (:7.0.5) to keep a fleet on identical code. Install and DSN/driver setup for each backend are in docs/INSTALL.md; the agent-facing contract is in AGENTS.md.
Quick Start
🚀 Fastest path — the free-backend Quickstart. Two commands to durable, MCP-accessible agent memory on a free database. It’s built around IBM Db2 12.1 — the reference deployment for the “mnemos on Db2” IBM TechXchange write-up — but the exact same image and steps run unchanged on Oracle AI Database 26ai Free, PostgreSQL + pgvector, or MariaDB 11.7+. Pick a backend in
quickstart/docs/BACKENDS.md; we lead with Db2,
安装
docker run -p 5002:5002 -v mnemos-data:/data ghcr.io/ncz-os/mnemos-enterprise:latest配置
{
"mcpServers": {
"mnemos": {
"command": "mnemos",
"args": ["serve", "mcp-stdio"],
"env": {
"MNEMOS_BASE": "http://<host>:5002",
"MNEMOS_API_KEY": "<token>"
}
}
}
}