Hermes Workflows

What Hermes Workflows does

You build a workflow in a visual editor based on @xyflow/react, validate it, preview the compiled Hermes plan, then press Play to run it with live per-node progress. There are seven node types: agent_task, prompt, script, condition, human_review, wait and finish. Each node compiles to a native Hermes primitive, meaning Kanban cards, Cron jobs and Profile assignments, with results sent through the host delivery router and skills taken from the host catalog.

Triggers are manual, cron, or event-based through webhook, github or api. An input_mapping setting passes a prior node's output to a later node, and the engine fails loudly if an expected output never appears. Workflows export to YAML or JSON and can be re-imported on a clean install. Each agent node reports duration, tokens, tool calls and errors, and pending dangerous-command approvals show in the run inspector. Release-oriented runs can resume failed runs and drive adopted board cards one at a time on a shared feature branch.

Key features

• Visual graph editor with validation and a compiled-plan preview
• Seven node types including human_review and wait
• Manual, cron and event triggers
• Run advancement through native Kanban lifecycle hooks, with a cron tick as a safety net
• YAML and JSON export and import of workflow specs
• Per-node telemetry and a run inspector with approval prompts

When to use it

• Chain agent tasks and shell steps with a human review gate in the middle
• Wait for an external signal such as a merged pull request before continuing
• Move a workflow between Hermes installs as a portable spec

Who it is for: Hermes Agent users who need automations with several steps, branches and approvals rather than a single scheduled prompt.

How it fits with Hermes Agent

Built for Hermes Agent as a dashboard plugin whose nodes compile to Hermes Kanban, Cron and Profile features.

FAQ