Hermes Atlas
Multi-agent & orchestration

Hermes Multi-Agent Research System

pedicelmarketing/hermes-multi-agent

Layered research architecture combining Paperclip, Hermes and MemOS with self-improving feedback loops

In short

Hermes Multi-Agent Research System is a layered multi-agent research and execution architecture that combines Paperclip, Hermes Agent and MemOS with autoresearch-style self-improving feedback loops.

What Hermes Multi-Agent Research System does

This repository is a glue layer that orchestrates external services rather than vendoring them. A patched MemOS fork supplies memory with Neo4j and Qdrant dependencies, the Hermes Agent CLI is installed from the official script, and setup-web-stack.sh bootstraps Firecrawl for search and scraping, SearXNG for meta-search through a docker-compose override file, and the Camofox anti-bot browser on port 9377.

The layout separates deploy/, the clean install source; skills/, dev copies of all skills shared into ~/.hermes/skills/ through external_dirs; memos-setup/learnings/, session notes and audit reports; and tests/, six-part blind audit prompts with remediation reports. A CLAUDE.md holds the canonical working rules. The web stack deliberately tracks upstream main for Firecrawl and the latest SearXNG image instead of pinning commits, trading occasional breakage for current anti-bot capability. The project includes research skills, architecture documents and session state for the Paperclip, Hermes and MemOS layers.

Key features

  • Paperclip, Hermes and MemOS layered architecture
  • Autoresearch-style self-improving feedback loops
  • Web stack bootstrap for Firecrawl, SearXNG and Camofox
  • Skills shared into ~/.hermes/skills/ through external_dirs
  • Blind audit prompts and remediation reports in tests/
  • Deploy installer for a fresh machine

When to use it

  • Standing up a Hermes research environment with its own memory and web search stack
  • Studying a worked example of Hermes, Paperclip and MemOS working together
  • Reusing the research skills and audit prompts in another Hermes setup

Who it is for: Advanced users building multi-agent research systems on Hermes Agent with self-hosted memory and web tools.

How it fits with Hermes Agent

Built around Hermes Agent: it installs the Hermes CLI and shares its skills into ~/.hermes/skills/, with Hermes as one of the three layers.

How to install Hermes Multi-Agent Research System

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

pip install git+https://github.com/sergiocoding96/memos
cd ~/.memos && bash start-memos.sh &
git clone https://github.com/sergiocoding96/hermes-multi-agent
cd hermes-multi-agent/deploy && ./install.sh

Requirements: A patched MemOS install, the Hermes Agent CLI, and Docker for the Firecrawl and SearXNG web stack

Note: The README flags an active migration (Sprint 2) from a MemOS server model to an embedded plugin model, and the repository has no license file.

FAQ

What is Hermes Multi-Agent Research System?

Hermes Multi-Agent Research System is a layered architecture that combines Paperclip, Hermes Agent and MemOS for research and execution. It adds autoresearch-style feedback loops that improve the system over time.

How do I install Hermes Multi-Agent Research System?

The README's fresh-machine quick start installs a patched MemOS and starts it, installs the Hermes CLI, runs deploy/install.sh from a clone, and then runs setup-web-stack.sh to bootstrap Firecrawl, SearXNG and Camofox. The deploy/README.md has the full reference.

Does Hermes Multi-Agent Research System work with Hermes Agent?

Yes. Hermes Agent is one of its three layers, and the repository's skills are shared into the Hermes skills directory.

Similar multi-agent for Hermes Agent

All multi-agent

Related guides: Build agent teams with the Hermes kanban board · Run multiple Hermes agents with profiles · Connect Hermes agents on several machines with Hermes Desktop