Video: "Hermes Agent Kanban Swarms AI SEO is INSANE!" by Julian Goldie on YouTube.
Why SEO is a good fit for swarm agents
An SEO project decomposes naturally into parallel subtasks. Keyword research does not need to wait for the content brief. Internal link planning does not need to wait for keyword research. These are largely independent jobs — which makes them well-suited to a model where multiple agents work simultaneously from a shared task board.
The Kanban setup in Hermes matches this. You create a board with columns for each stage, populate it with tasks, and let the dispatcher assign specialist agents to each one. A research agent, a content brief agent, and a technical SEO agent can all be live at the same time working on different cards.
What the swarm workflow looks like on a real project
Julian Goldie ran this against an actual site — feeding in a target keyword set and letting the swarm work through the board. The research agent pulled competitor SERP data. The brief agent structured each piece against topical authority clusters. A separate agent handled schema and meta tags. Total wall-clock time for a 20-page site plan came down substantially compared to running each step in sequence.
That said, the quality at each step reflects the quality of the agent profile behind it. A vaguely defined brief agent produces vague briefs. The work is in the setup, not the running.
The SEO-specific advantages
SEO work involves a lot of repetitive structure. Writing 40 meta descriptions, building 40 H1s, checking 40 pages for thin content — these are tasks where the pattern is fixed and the volume is the problem. Swarm agents handle volume well. You define the task once, put 40 cards on the board, and let the agents work through them.
There is also a practical benefit for agencies. You can run the same board structure across multiple client sites, swapping out the target domain and keyword set but keeping the agent profiles consistent. That reduces the time spent rebuilding the same workflow each engagement.
Where it still needs oversight
The inter-agent handoff is where things can break. If the research agent outputs a format the brief agent does not expect, the workflow stalls at that join. Julian flagged that you need to test handoffs carefully before running anything on a client site — the agents are not reading each other's minds, they are just passing files.
LLM-generated SEO content also still needs a human pass before it goes live. The swarm accelerates production; it does not replace editorial judgement. Factual errors, tone mismatches, and thin sections still slip through even on well-configured runs.
Where this connects to NordSys
Multi-agent SEO workflows are genuinely useful — but only when the agent profiles are built around your actual strategy and your site's specific context. Generic setups produce generic output. We build Hermes Agent configurations that reflect how a real SEO project should run: what research the brief agent needs, how the content agent should structure its output, what the handoff format looks like at each step. Our SEO & AI Ranking service covers the full picture.
See our SEO & AI Ranking service →