Human Memory
konglong87/human-memory
Memory skill modeled on human cognition with six memory layers, decay and risk warnings
Human Memory is a skill and Python library that gives AI agents a six-layer memory modeled on human cognition, with importance scoring, forgetting curves and warnings drawn from past lessons. It stores everything in SQLite and is tagged as a Hermes skill.
What Human Memory does
The six layers are short-term memory, episodic memory, long-term memory, world knowledge, lessons learned and a running log. Memories are scored on intensity, duration, danger and scarcity, strengthened when triggered again, and faded over time with an exponential clarity formula. A graph engine links entities, and a warning engine detects scenarios such as driving in rain and matches them with stored lessons at four levels, P0 to P3, from life-threatening interrupts to reference queries.
Storage is embedded SQLite with no external database. The repository installs as a Python package and can also run as a Claude Code plugin with a SessionStart hook, offering commands such as /human-memory write, query, warning and stats. Industry knowledge can be added as YAML files in a world-knowledge folder. The README documents Claude Code setup in detail; its link to Hermes is the hermes-skill topic tag.
Key features
- Six memory layers from short-term to lessons learned
- Importance scoring with reinforcement and exponential decay
- A graph engine for entity relationships
- A warning engine with four levels, P0 to P3
- Embedded SQLite storage with no external database
- A Claude Code plugin with write, query, warning and stats commands
When to use it
- Giving an agent persistent lessons learned from past mistakes
- Warning about a risky scenario when it matches a stored lesson
- Building a personal or team knowledge store with forgetting built in
Who it is for: Developers who want an experimental, cognition-inspired memory layer for AI agents.
How it fits with Hermes Agent
The repository is tagged hermes-skill and also lists Claude Code, Cursor and OpenClaw, but its README documents setup only for Claude Code and as a Python package.
How to install Human Memory
These commands are copied from the project's README. Check the repository for the latest steps before you run them.
git clone https://github.com/konglong87/human-memory.git
cd human-memory
pip install -r requirements.txt
pip install -e .FAQ
What is Human Memory?
Human Memory is a memory system for AI agents modeled on human cognition. It records experiences and lessons, scores their importance, lets them fade over time and raises warnings when a similar scenario returns.
Does Human Memory work with Hermes Agent?
The repository is tagged hermes-skill and also names Claude Code, Cursor and OpenClaw. The README documents setup only for Claude Code and as a Python package, with no Hermes-specific steps.
Is Human Memory free and open source?
Human Memory is open source under the MIT license.
Similar memory for Hermes Agent
All memoryExabase M-1 long-term memory provider plugin for Hermes Agent
hermes-pro Mem0 for HermesMem0 memory plugin for Hermes Agent that routes extraction through Hermes's auxiliary model
EESIZ HermesYumeNightly sleep-style memory consolidation for Hermes Agent, with long-term facts in LanceDB
goldenSniperOS Hermes RecallEngram-compatible memory provider for Hermes Agent with automatic recall and capture on every turn
hermegeddon Hermes Memory WikiHermes plugin for a persistent markdown knowledge vault with claims, lint checks and QMD semantic search
vokasug brain-vectorHermes skill for semantic search over a personal knowledge base using LanceDB and local MLX embeddings
Related guides: SOUL.md for Hermes Agent: what it is and how to write one · Run multiple Hermes agents with profiles