An SEO Plan Written by a Crew of AI Agents, Not One Prompt
Editor's SEO agent runs a multi-agent crew covering technical SEO, content strategy, and keyword gaps together, instead of one prompt trying to cover all three at once.
The AI-SEO tool space right now is mostly single-purpose: one tool for keyword research, another for content briefs, another for technical audits, and a person stitching the outputs together into something that actually resembles a plan. Editor's SEO agent takes a different shape — a multi-agent system that produces the plan as one coordinated output, rather than a single prompt trying to cover technical recommendations, content strategy, and keyword gaps all at once and doing a mediocre job of each.
What "crew" means here, specifically
The implementation lives in frontend/apps/editor/src/modules/seo-agent/crew.ts, and its own header describes it directly as a "multi-agent system for comprehensive SEO analysis and planning." It works from a defined input structure and produces a structured SEOPlan output covering three distinct areas: technical SEO recommendations, content strategy, and topic, keyword, and content-gap analysis. Those are genuinely different kinds of analysis — technical SEO is about crawlability, structure, and site health; content strategy is about what to write and why; keyword and gap analysis is about where the actual opportunity is relative to what's already ranking. A single prompt asked to do all three at once tends to produce a document that's shallow across the board rather than solid in each area. A crew structure — where different parts of the analysis are handled as distinct concerns and then assembled — is built specifically to avoid that shallowness.
Why this beats chaining single-purpose tools yourself
The alternative most people are doing today is manual: run a keyword tool, run a technical audit tool, run a content-brief generator, then sit down and reconcile the three outputs into something coherent yourself. That reconciliation step is real work, and it's exactly the step a built-in multi-agent planner is meant to remove — the different analyses are produced with the same underlying context and assembled into one plan, rather than three separate documents you have to cross-reference by hand.
What to know before leaning on it fully
The crew is routed through Vercel AI Gateway, and there's a fallback.ts sitting alongside the crew implementation, which suggests a degraded-mode path exists for when the primary flow can't complete cleanly. That's worth knowing rather than assuming: a fallback path existing is a sign the system is built to handle failure gracefully rather than simply erroring out, but exactly how degraded a fallback result is compared to the full crew's output is worth checking directly if you're depending on the plan's completeness for something high-stakes, rather than assuming every plan you get back reflects the full three-part analysis running cleanly.
The actual pitch
Not "an AI that knows SEO" — that's the pitch every competitor in this space is already making with a single well-tuned prompt. The distinct claim here is structural: a plan assembled from purpose-built analysis across the three areas that actually make up an SEO plan, produced together rather than stitched together by you afterward.