convez.ai
MCP ServerDevOps & InfraestruturaAvançado
ncz-os

MNEMOS — Sistema Operacional de Memória para Agentes IA

ncz-os/mnemos

Memory OS production-grade para agentes IA com persistência multi-backend

Ver no GitHub
BomPluginScore · de 0 a 100

Runtime de produção para gerenciamento completo do ciclo de vida da memória de agentes IA, com persistência multi-backend (SQLite, PostgreSQL, Oracle, IBM Db2), busca vetorial, compressão, versionamento e raciocínio integrado. Compatível com Claude, LangChain, CrewAI, AutoGen e outros via MCP e gateway OpenAI-compatível.

Instalação

$npx convez-ai ncz-os/mnemos

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
35
Forks
7
Acompanhando
1
Saúde
75%
Último commit
há 4 dias
Versão
v6.1.7
Issues abertas
3
Contribuidores
8

Casos de uso

  • Armazenamento persistente e recuperação de contexto para agentes autônomos
  • Federação de memória entre múltiplos agentes e frameworks
  • Compressão e raciocínio sobre histórico de interações de longa duração
  • Auditoria, versionamento e conformidade GDPR em memória de agentes

Veredito

MNEMOS é uma solução enterprise-ready para gerenciar memória em ambientes de agentes IA complexos. Arquitetura modular, multi-backend e interoperável, mas requer expertise significativa em deployment e operação de sistemas distribuídos.

README

A tradução para o português deste README ainda está na fila.

MNEMOS + GRAEAE

MNEMOS v6.0.0 is the memory operating system for serious agentic work: a packaged FastAPI runtime, EPIMONE — the four-backend persistence layer (SQLite + sqlite-vec by default, PostgreSQL + pgvector, Oracle Database 26ai HNSW INMEMORY NEIGHBOR GRAPH, IBM Db2 12.1.5 (EAP) DiskANN vector), GRAEAE reasoning bus, operator-audited compression stack, divergent dream-state pipeline (REPLAY -> CLUSTER -> CONSOLIDATE -> SYNTHESISE -> EXTRACT), GDPR right-to-be-forgotten worker, PERSEPHONE archival subsystem, PANTHEON unified LLM facade, KRONOS recall observability, and CLI-first deployment surface.

MNEMOS 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.

v6.0.0 — the split-distribution release. MNEMOS is now a small core (mnemos-core) plus separately installable mnemos.* namespace subsystems (GRAEAE, PANTHEON, KNEMON, CHARON) and the standalone STIPHOS hive service. It adds Oracle Database 26ai and IBM Db2 12.1.5 (EAP) as first-class persistence backends alongside PostgreSQL and SQLite. Turnkey container images are published to ghcr.io/ncz-os (mnemos-core, mnemos, mnemos-enterprise, mnemos-stiphos); see docs/INSTALL.md and AGENTS.md. Enterprise backend driver, DSN, and migration steps are in docs/INSTALL.md. Development history continues on the feat/oracle-port branch.

Quick Start

Memory and reasoning runtime for AI agents: persistent search, versioned storage, webhook fanout, and a unified LLM routing bus - all behind a single MCP interface.


1. Agent-driven install

Paste into Claude Code, Cursor, or Codex. The agent runs the install; you confirm. Agents should read AGENTS.md — it has the machine-readable module registry and a deterministic procedure for installing exactly the requested modules on the operator's arch + backend.

The pip package is mnemos-core (the subsystems are separate dists pulled via extras). mnemos is the published image name, not a pip package.

Turnkey (container, any arch):

docker run -p 5002:5002 -v mnemos-data:/data ghcr.io/ncz-os/mnemos:latest
# everything image: core + graeae + pantheon + knemon + charon. SQLite by default.
# Point at a real DB with -e MNEMOS_DATABASE_DSN='postgres://…' (or oracle://… thin).

pip (compose your own):

Install MNEMOS on this machine.

Steps:
1. pip install 'mnemos-core[server]'   # everything; arch-neutral (no openvino)
2. mnemos init                         # scaffold config + token
3. mnemos serve                        # start API on :5002
4. mnemos doctor                       # verify subsystems
5. Set MNEMOS_BASE=http://localhost:5002 and MNEMOS_API_KEY=<token from step 2>
   in shell env and any agent config that needs to reach it.

Edge device (SQLite kernel only): pip install 'mnemos-core[edge]'
Single subsystem, e.g. reasoning: pip install 'mnemos-core[graeae]'
Hive (STIPHOS) is a SEPARATE service: pip install 'mnemos-stiphos[mcp]' (port 8080)

Enterprise backends (Oracle Database 26ai, IBM Db2 12.1.5 EAP). Turnkey is the amd64-only mnemos-enterprise image (everything + Oracle/Db2/ MySQL drivers baked in). Note: Oracle uses the thin driver, so it also runs on the plain mnemos image and on arm64 — only Db2 actually requires enterprise. See docs/INSTALL.md for full driver, DSN, and migration steps.

# Turnkey (amd64):
docker run --platform linux/amd64 -p 5002:5002 \
  -e MNEMOS_DATABASE_DSN='db2://user:pass@host:50000/dbname' \
  ghcr.io/ncz-os/mnemos-enterprise:latest

# Or from source:
git clone https://github.com/ncz-os/mnemos && cd mnemos
python -m pip install -e '.[server,enterprise]'   # or '.[server,oracle]' / '.[server,db2]'
export MNEMOS_DATABASE_DSN='oracle://user:pass@host:1521/service_name'
# or:  MNEMOS_DATABASE_DSN='db2://user:pass@host:50000/dbname'
mnemos install --profile server
mnemos serve --profile server

2. Connect an agent via MCP

Add to ~/.claude/mcp_servers.json (Claude Code) or equivalent:

{
  "mcpServers": {
    "mnemos": {
      "command": "mnemos",
      "args": ["serve", "mcp-stdio"],
      "env": {
        "MNEMOS_BASE": "http://<host>:5002",
        "MNEMOS_API_KEY": "<token>"
      }
    }
  }
}

For HTTP/SSE transport (ChatGPT, remote agents): mnemos serve mcp-http on :5004.

Key MCP tools the agent gets:

ToolWhat it does
search_memoriesSemantic + filtered search across the memory store
create_memoryWrite a new memory with category, tags, and content
get_memoryFetch a memory by ID
kg_searchQuery the knowledge-graph triple store
kronos_anomaliesSurface recall anomalies and memory health signals
list_deletionsList soft-deleted memories pending hard deletion

3. Webhooks + integrations

IntegrationWhat connectsHow
Claude CodeHooks fire on session-start, prompt-submit, stop - auto-log to MNEMOSintegrations/claude-code/ - copy hooks + set MNEMOS_BASE
ZeroClawZeroclaw agent reads/writes memories via MCPintegrations/zeroclaw/ + mnemos serve mcp-stdio in zeroclaw config
OpenClawOpenClaw gateway routes memory ops through MCPintegrations/openclaw/ + MCP server entry in openclaw.json
HermesOptional memory skill mounts MNEMOS as a tool providerintegrations/hermes/optional-skills/memory/mnemos/
Webhooks (any)Push memory.created, memory.updated, memory.deleted, consultation.completed events to any HTTPS endpointPOST /api/webhooks/register with {"url": "...", "events": [...]}
Cursor / Cline / Continue.dev / Zed / AiderAny MCP-capable IDE connects via stdio or HTTP transportSee docs/connectors/

Full documentation: docs/

Architecture

MNEMOS is a packaged FastAPI service with a single mnemos CLI for installation, serving, MCP transport, and operational checks. Agents connect through MCP stdio, MCP HTTP/SSE, REST, or OpenAI-compatible SDKs, while the runtime routes memory, reasoning, session, webhook, federation, portability, and observability work through the mnemos/ package. Persistence is selected by profile and DSN: SQLite + sqlite-vec for edge and development installs, PostgreSQL + pgvector for server deployments, Oracle Database 26ai (23.26.1-ee, HNSW INMEMORY NEIGHBOR GRAPH, JSON Duality, TDE) for enterprise installs, and IBM Db2 12.1.5 (native VECTOR(768, FLOAT32) + DiskANN vector index; runs through Db2 Oracle Compatibility Mode with cursor-level Oracle→Db2 token translation — a native Db2 dialect port is on the v6.x roadmap, see docs/v6.1-roadmap.md. Db2MemoryRepository.semantic_search emits native Db2 SQL — VECTOR_DISTANCE(..., EUCLIDEAN) + FETCH APPROX FIRST — engaging the DiskANN index on the user-facing query path) for enterprise installs. All four backends implement the same PersistenceBackend ABC (mnemos/persistence/base.py) and share tests/test_persistence_parity.py. GRAEAE handles multi-provider reasoning and model routing; MOIRAI handles operator-audited compression through APOLLO and ARTEMIS.

Documentation

TopicFile
Installationdocs/INSTALL.md
Specificationdocs/SPECIFICATION.md
System requirementsdocs/SYSTEM_REQUIREMENTS.md
Memory architecturedocs/MEMORY_ARCHITECTURE.md
Compressiondocs/COMPRESSION.md
GRAEAE reasoningdocs/GRAEAE_FEATURES.md
PANTHEON provider facadedocs/PANTHEON.md
KRONOS observabilitydocs/KRONOS.md
Portability formatdocs/MEMORY_EXPORT_FORMAT.md
Scalingdocs/SCALING.md
Single-binary buildsdocs/SINGLE_BINARY.md
Operationsdocs/OPERATIONS.md
Benchmark harnessscripts/bench_v4.py — cross-backend vector-search harness (PG / Oracle / Db2 / SQLite). Results published post-GA.

License

MNEMOS is licensed under the Apache License, Version 2.0. See LICENSE for the full text.

Ver o original no GitHub

Bom para

Relacionados