MNEMOS
ncz-os/mnemos
Memory operating system for AI agents with MCP, OpenAI-compatible and REST interfaces
MNEMOS is a memory operating system for AI agents that manages the lifecycle of agent memory and exposes it through MCP, an OpenAI-compatible gateway and native /v1 REST endpoints. Hermes is among the frameworks it lists as interoperable.
What MNEMOS does
MNEMOS packages a FastAPI runtime with a CLI-first deployment surface. Its EPIMONE persistence layer supports six backends: SQLite with sqlite-vec by default, PostgreSQL with pgvector, Oracle AI Database 26ai, IBM Db2 12.1, MySQL 9.0 Enterprise/HeatWave and MariaDB 11.7 or later, and each backend creates its schema on first connect. The GRAEAE reasoning bus and PANTHEON LLM facade sit alongside the memory store.
Other subsystems cover a dream-state pipeline (replay, cluster, consolidate, synthesise, extract), a GDPR right-to-be-forgotten worker, PERSEPHONE archival, CHARON import and export in MIF and MPF formats with STYX encrypted backups, and KRONOS recall observability. The pip package is mnemos-core, with GRAEAE, PANTHEON and KNEMON as separate installable distributions, and the container image is ghcr.io/ncz-os/mnemos-enterprise. The authoritative source is on GitLab, and this GitHub repository exists to host that image.
Key features
- Six persistence backends with SQLite and sqlite-vec by default
- MCP, OpenAI-compatible gateway and native /v1 REST interfaces
- Dream-state memory pipeline from replay to extraction
- GDPR right-to-be-forgotten worker
- MIF and MPF import and export with encrypted STYX backups
- Multi-arch container image with a single docker run quick start
When to use it
- Giving agents such as Claude Code, OpenClaw or Hermes a shared persistent memory service over MCP
- Running agent memory on an existing PostgreSQL, Oracle, Db2, MySQL or MariaDB database
- Exporting and migrating agent memory between memory systems
Who it is for: Developers and operators who need a self-hosted memory service shared across several agent frameworks.
How it fits with Hermes Agent
Hermes appears in its list of interoperable frameworks reached through MCP, the OpenAI-compatible gateway and REST; the opening part of the README gives no Hermes-specific setup.
How to install MNEMOS
These commands are copied from the project's README. Check the repository for the latest steps before you run them.
docker run -p 5002:5002 -v mnemos-data:/data ghcr.io/ncz-os/mnemos-enterprise:latestNote: The authoritative source lives on GitLab (gitlab.com/ncz-os/mnemos), and this GitHub repository is retained only to host the mnemos-enterprise container image.
FAQ
What is MNEMOS?
MNEMOS is a memory operating system for agentic AI. It manages the lifecycle of agent memory (write, embed, search, compress, version, audit, export and import) behind MCP, an OpenAI-compatible gateway and /v1 REST endpoints.
Does MNEMOS work with Hermes Agent?
Hermes is listed among the frameworks MNEMOS interoperates with, alongside Claude Code, OpenClaw, LangChain and others, through MCP, the OpenAI-compatible gateway and REST. The opening part of the README does not describe a Hermes-specific setup.
How do I install MNEMOS?
Run the container with docker run -p 5002:5002 -v mnemos-data:/data ghcr.io/ncz-os/mnemos-enterprise:latest, which uses SQLite by default. Set MNEMOS_DATABASE_DSN to point at another database. The pip package is mnemos-core, and mnemos is only the image name.
Similar memory for Hermes Agent
All memoryFilesystem memory for Hermes Agent built on markdown and YAML bundles in Open Knowledge Format
elkimek Hermes + Honcho + GBrainGuide to running Honcho and GBrain together with Hermes Agent for relationship and source memory
smarzola Hermes Local MemoryLocal-first SQLite memory provider for Hermes Agent with peers, cards, search and consolidation
M4F-S GomaaLocal-first hierarchical memory engine with Obsidian vault storage, hybrid search and an MCP server
TJCurnutte Memory HiveFile-backed shared memory for AI agent teams with private silos, a shared hive and a curator workflow
r0b0tlab r0b0tlabbra1nFilesystem-first Markdown and Obsidian memory vault for Hermes with SQLite FTS5 search and secret scanning
Related guides: SOUL.md for Hermes Agent: what it is and how to write one · Run multiple Hermes agents with profiles