Memory OS: memory tool for Hermes Agent

Seven-layer local memory system for Hermes Agent with Qdrant, structured facts and a curated wiki

What Memory OS does

Memory OS adds seven memory layers to Hermes Agent so that it keeps context between sessions. Layer 1 is workspace files, MEMORY.md, USER.md and CREATIVE.md, injected into the system prompt every turn. Layer 2 is session history in state.db with SQLite and FTS5 full-text search. Layer 3 holds structured facts in memory_store.db, with entity resolution and trust scores that a feedback loop adjusts over time. Layer 4 is cross-session recall through a heavily forked Icarus plugin, and the stack also includes a Qdrant vector database and an auto-curated wiki pipeline.

Context injection is selective, and an explicit Ground Truth hierarchy tells the agent to use the injected memory. The infrastructure runs locally with Docker services (Qdrant, Redis and an ARQ worker) and works with any LLM provider Hermes supports, including OpenRouter, OpenAI, Anthropic and Ollama. Version 0.2.0 added a one-command installer that sets up the Docker services, SQLite databases, the Icarus plugin and the environment, and smoke and ingestion tests ship with the repository.

Key features

Workspace files MEMORY.md, USER.md and CREATIVE.md injected into the system prompt each turn
SQLite and FTS5 full-text search across session history
Structured facts with entity resolution and trust scoring
Cross-session recall through a forked Icarus plugin and a Qdrant vector database
Auto-curated wiki pipeline and a Ground Truth hierarchy for injected memory
One-command installer plus smoke and ingestion tests

When to use it

Stop repeating project context at the start of every Hermes Agent session
Search every past conversation semantically and by keyword
Keep durable facts about your stack and projects with trust scores

Who it is for: Hermes Agent users who want deeper long-term memory that runs on their own machine.

How it fits with Hermes Agent

Built for Hermes Agent: it extends its MEMORY.md and USER.md workspace files and state.db sessions with additional memory layers and a plugin.

How to install Memory OS

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

curl -sSL https://raw.githubusercontent.com/ClaudioDrews/memory-os/main/setup.sh | bash

Requirements: Hermes Agent, Docker (Qdrant, Redis and an ARQ worker) and Python 3.11+

FAQ

What is Memory OS?

Memory OS is a seven-layer memory operating system for Hermes Agent. It combines workspace files, session search, structured facts, a vector database and an auto-curated wiki to inject the right context into each turn.

Does Memory OS work with Hermes Agent?

Yes, it is built for Hermes Agent and requires it. It works with any LLM provider Hermes supports, including OpenRouter, OpenAI, Anthropic and Ollama.

How do I install Memory OS?

Run curl -sSL https://raw.githubusercontent.com/ClaudioDrews/memory-os/main/setup.sh | bash. It sets up the Docker services, SQLite databases, the Icarus plugin and the environment, and the README keeps a manual guide as a fallback.