Hermes Agent
- Judgment, routing, and debug when the path is unclear
- Unstructured text, OCR review, and natural-language tasks
- Each call uses tokens
PetPetAgent
Product features
Three capabilities on one desk companion — see growth, tighten the docs that steer your Agent, and spend fewer tokens on waste.
Connect n8n automation
Hermes is good at judging. n8n is good at repeating the same steps without improvising. PetPetAgent runs both on your Windows desk: design workflows in the browser, then let the Agent call them when a job actually needs a stable pipeline.
If you already have an AI agent, why add n8n?
They are not substitutes. AI answers can change from run to run — that flexibility is useful, and it is also a risk when the job must be identical every time.
Fixed steps you can watch
Backups at 3 a.m., a Monday report, a file rename rule — you should not leave those to an agent’s improvisation. An n8n workflow is deterministic: you define each node, and it follows that path. No surprise branches, no “creative” detours.
Spend tokens only where judgment is needed
Every LLM call costs tokens. Format conversion, scheduled notices, and filing do not need a model. Put those on n8n and keep Hermes for unstructured data, exceptions, and decisions. The same month of work usually burns far fewer tokens.
A full execution log
An agent’s reasoning is hard to audit. n8n records when a run started, what each node received, what it emitted, and where it failed. When something breaks, you open the node — you do not guess from a chat reply.
Each side does a different job
n8n workflows
- Clear rules, timers, and system-to-system moves
- Structured data transform and hand-off
- Runs locally — no extra token bill for the pipeline itself
How they work together on PetPetAgent
Use n8n as the skeleton: nodes you can inspect, rerun, and trust. Put Hermes on the nodes that need a brain — or let Hermes start a whole workflow through MCP when you ask in chat. The Agent decides and debugs; n8n executes the stable path. Only exceptions come back to the Agent. That is the same split used in a strong agent architecture: judgment on one side, a fixed program on the other.
Local n8n, on your machine
The Automation workstation installs and starts n8n on localhost. Workflows and data stay on your PC. Open the editor in your browser, then return to chat so Hermes can list and run those workflows — without sending the pipeline to a third-party automation cloud.