The gist

Hermes Multi-Agent Workflow is an open-source template from Tonbi Studio for building autonomous multi-agent pipelines. It provides a reusable skeleton for workflows that source items, score them against a rubric, conduct research, and route them for fulfillment. The framework gives developers and their coding agents a clear, working structure to adapt for their specific domains, rather than building complex agentic systems from scratch. It comes with a pre-configured example pipeline for finding AI agent pain points.

What it does

Detect items from configurable sources and add them to a triage pipeline.
Deduplicate, score, and research items in parallel using autonomous agents.
Route items to different fulfillment paths based on a custom classification map.
Pause the workflow at a human approval gate for review before fulfillment.
Scaffold the required Kanban board setup for Hermes orchestration.

How it works

This is a self-hosted Python template that a developer adapts for a specific domain. The entire pipeline—sources, scoring rubrics, research tasks, and fulfillment paths—is defined in a central triage.yaml configuration file. The user runs the CLI to validate the configuration and execute the multi-agent workflow, which is orchestrated on a Hermes Kanban board. The project is open-source under an MIT license and requires a separate Hermes installation to run.

Best for

This template is best for developers building structured, multi-agent systems for tasks like content curation, lead scoring, or bug triage who want a domain-agnostic skeleton to start from.

Watch out for

This is a developer template, not a turnkey application. It requires a separate Hermes installation and significant configuration of profiles, authentication, and agent skills before it can be used for a live pipeline.