Agent Memory: memory tool for Hermes Agent

Agent Memory

hsuanshao/AgentMemory

Self-hosted Rust memory service with layered L0 to L3 memory and hybrid BM25 plus vector recall

In short

Agent Memory is a self-hosted memory service for AI agents, written in Rust, with layered memory and hybrid recall. Its topics include hermes-agent and hermes-plugin, and its docs cover a Hermes integration.

What Agent Memory does

The core service, memory-core, follows a clean architecture and stores data in SQLite and local files, or PostgreSQL. It organizes memory in four layers from L0 to L3, distills it asynchronously with an LLM, and recalls it with hybrid BM25, vector and RRF ranking. Team, agent and user isolation, skill memory, and asset and ACL metadata are part of the design.

Any OpenAI-compatible endpoint, including self-hosted models, can serve as the LLM and embedding provider. The installer can instead use the Claude Code, Antigravity or Codex CLI. A TypeScript Monitor Panel is included, and a v3 RPC API is exposed on port 8420.

The README says remote clients cover Claude Code, Codex, Antigravity and Hermes integrations, installed with the amem client. Details are in docs/REMOTE-CLIENTS.md.

Key features

L0 to L3 layered memory with asynchronous distillation
Hybrid BM25, vector and RRF recall
Team, agent and user isolation
SQLite by default, or PostgreSQL with optional Redis
Any OpenAI-compatible LLM or embedding endpoint
TypeScript Monitor Panel

When to use it

Running a shared memory server for several agents on different machines
Keeping agent memory on your own hardware with a local model
Reviewing stored memory through the Monitor Panel

Who it is for: Developers who want a self-hosted memory backend shared by coding agents.

How it fits with Hermes Agent

It is a general memory service with a documented Hermes integration, and its topics are tagged hermes-agent and hermes-plugin.

How to install Agent Memory

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

./install.sh

Requirements: Docker with Docker Compose; an OpenAI-compatible LLM endpoint or the Claude Code, Antigravity or Codex CLI

FAQ

What is Agent Memory?

Agent Memory is a self-hosted memory service for AI agents, written in Rust. It provides layered memory, asynchronous distillation and hybrid recall.

Does Agent Memory work with Hermes Agent?

Yes. The README lists a Hermes integration among its remote client integrations, and the repository is tagged hermes-agent and hermes-plugin.

How do I install Agent Memory?

Run ./install.sh from the repository. It asks about the storage backend, coordination backend and LLM, then starts the services.

Is Agent Memory free and open source?

Yes. It is released under the MIT license.