Nod is n8n’s agent that browses as you and asks before it acts. I already run an agent that browses as me and asks before it acts. So do I need Nod?

The interesting question isn’t whether Nod is good. Its approval flow and desktop-VM packaging are cleaner than anything most people cobble together by hand. The real question is whether a second agent holding your logins earns its keep when one already does.

What Nod Claims to Do

Nod is an Apache-2.0 personal agent from n8n Labs. You can run it as a desktop Linux VM or spin it up as a Docker stack — PostgreSQL, the agent runtime, a workspace container, and a browser container sitting behind a firewall. The pitch is simple: an unattended helper that lives alongside your daily work, handles routine chores, browses the web in a secure container, stores your passwords in an encrypted vault, and prompts you whenever it hits something it isn’t sure about.

It leans hard on prepackaged convenience. It comes with a prebuilt catalogue of SaaS integrations (Gmail, Slack, Notion, Drive, Calendar) wired up via OAuth, a dedicated UI for managing credentials, and an approval prompt that handles unknown requests without you writing custom code.

What Hermes Actually Is

I do all of that on Hermes, but the architecture is entirely different. Hermes isn’t a single monolithic application or a ready-made VM. It’s a general-purpose runtime where agents run as side-by-side profiles. Each profile is simply a directory of state — config files, credentials, memory notes, session history, skills, and scheduled jobs.

When you compare what Nod claims to do with what Hermes achieves, the capability gap disappears. The configuration gulf, however, is massive.

1. Automation & Triggers: Nod runs scheduled tasks. In Hermes, scheduled jobs are straightforward prompts tied to cron schedules, dropping their output directly into whatever chat channel you attach. Inbound webhooks take POST requests, verify HMAC signatures, and turn payloads into agent prompts without human intervention.
2. Tools & Integrations: Nod uses Composio for SaaS connectors. Hermes uses Model Context Protocol (MCP) servers — point the agent at a server and its tools show up alongside built-in capabilities. For services like Gmail or Slack, Hermes uses hosted or local connectors requiring a one-time OAuth authorization.
3. Browsing & Desktop Interaction: Nod runs a browser container. Hermes splits browser work into two deliberate tools: Python-driven Chromium automation for headless tasks, and a separate secure credential vault (browser_vault) that fills logins without exposing passwords to model turns. When pages break under headless scripts, computer use drives a real desktop directly via accessibility trees and mouse/keyboard clicks.
4. Delegation & Document Production: A parent Hermes agent can spawn child agents with isolated terminal sessions and narrowed toolsets to split up heavy research. Document production handles file reading, patching, spreadsheet generation, and headless PDF renders directly to disk rather than spamming your chat window.
5. Memory & Procedural Skills: Hermes separates short curated memory notes (loaded into every session) from full-text vector search across past conversations. Skills are plain text manuals stored on disk, loading only when relevant so instructions don’t bloat the active context window.

Validating the Conclusion: More Powerful, Far Harder

The core thesis is straightforward: Hermes is vastly more powerful, but far harder to configure.

Nod wins on packaging and speed. If you want a zero-config approval prompt, a polished web vault UI for your passwords, and instant OAuth connectors to Gmail and Slack wrapped in a neat Docker compose stack or desktop app, Nod delivers that without you touching a line of config.

Hermes gives you absolute control and endless extensibility, but you pay for it in setup sweat. There is no graphical installer or prebuilt SaaS dashboard. If you want approval gating, you write hook scripts and set up allowlists yourself. If you want custom tool routing, you configure MCP servers and session profiles from scratch. If you want secure browser automation, you manage vault files and Python helpers. Nod doesn’t do anything Hermes fundamentally can’t; it just packages those capabilities into opinionated defaults.

In practice, that means hours of plumbing. Setting up Nod took a Docker compose file and an OAuth clickthrough. Getting Hermes to the same level of convenience meant learning its MCP surface, writing custom hook scripts for file-write gates, and building out session profiles before a single cron job ran unattended. That upfront setup cost is the tax you pay for sovereignty.

If you want a turnkey personal agent with a polished UI and immediate SaaS integrations, Nod is the pragmatic pick. If you want a general-purpose, deeply extensible agent runtime that can reach anywhere, parse any local file, delegate to sub-agents, and run unattended cron workflows — and you don’t mind wiring the plumbing yourself — Hermes remains unmatched.

Hermes does the work. Nod packages it.