20 Claude Code Agent Teams for Marketing, SEO, and Growth (2026 Ranked)

By Rakshit Yadav (@yadavrakshit60)•Oct 2026•11 min read

Short answer: a Claude Code "marketing agent team" is only as useful as the division of labor behind it — one seo-specialist that never writes full articles, one content-marketer that drafts them, one growth-strategist that decides where to spend the next week, and roughly 17 more roles that each own one output and hand off to the next. Below are 20 ranked agent/skill combos built for a solo SaaS founder's actual week — not a generic "AI marketing team" prompt pack — each mapped to a real role from AgenticKit's 34-agent setup, with what it produces and when to delegate it versus do it yourself.

Why a prompt library isn't a team

Most "Claude Code marketing agent" posts ship a list of prompts — "SEO audit prompt," "LinkedIn post generator," "cold email prompt" — and call it a team. That's a prompt library, not a team, and it breaks down the same way a single giant engineering subagent does: no ownership boundaries, no handoffs, and no way to tell which agent should actually run for a given task.

A real team has five things a prompt list doesn't:

  1. Scoped responsibility. positioning-strategist defines the message house; it doesn't write landing page copy. copywriter writes the copy; it doesn't decide the positioning. One founder running both roles in a single ChatGPT thread ends up with copy that drifts from the positioning within a week, because nothing is forcing consistency.
  2. A trigger that routes correctly. Each agent's frontmatter description has to be specific enough that "write a LinkedIn post" routes to social-media-strategist and not content-marketer. Vague descriptions mean everything routes to whichever agent you invoked most recently.
  3. A real handoff artifact. growth-strategist producing an experiment backlog is only useful if launch-coordinator and copywriter actually consume it — as a file, not a paragraph that gets lost in chat history.
  4. A "does not" boundary. product-marketer explicitly does not run multi-channel launches (that's launch-coordinator's job) and does not write sales outreach (that's sales-closer's). Without that line, the same agent ends up doing both badly.
  5. Honest outputs. customer-marketer and video-producer are built to never invent quotes, metrics, or testimonials — a constraint worth having for any marketing agent you let touch customer-facing copy.

Solo founders don't need 34 generalist prompts. They need the five or six roles below that cover SEO, growth, lifecycle, content, and support — plus the rest of the list for when the workload actually calls for it.

The 5 roles every SaaS founder's team needs first

RoleOwnsHands off to
seo-specialistKeyword strategy, page briefs, schema specscontent-marketer, react-specialist
growth-strategistAARRR funnel diagnosis, experiment backlogcopywriter, launch-coordinator, positioning-strategist
content-marketerEditorial calendar, long-form draftsseo-specialist
email-marketerLifecycle sequences, onboarding dripsanalytics-engineer
support-engineerTicket triage, repro steps, customer repliesdebugger (for stack-trace bugs)

If you install nothing else, these five cover the loop most solo founders run weekly: find the keyword, write the page, diagnose why signups stall, keep trial users onboard by email, and answer the tickets that come in while you do it.

The ranked list

#Agent/skillWhat it producesWhen to delegate vs. DIY
1seo-specialistKeyword strategy, pillar/cluster architecture, page briefs, JSON-LD specsDelegate for any new page or schema change; DIY only for one-off meta tweaks
2growth-strategistAARRR funnel map, growth loops, channel/experiment priorityDelegate when activation or retention has stalled; DIY for a single quick experiment idea
3copywriterLanding page copy, headlines, CTAs, ad copyDelegate for anything customer-facing; DIY for internal docs
4positioning-strategistCategory, competitive alternatives, message houseDelegate before any major repositioning or new pricing tier; DIY rarely — this one needs discipline
5content-marketerEditorial calendar, long-form articles, lead magnetsDelegate for scheduled content; DIY for a single reactive post
6social-media-strategist30-day LinkedIn calendar, hook library, post draftsDelegate for cadence; DIY for a one-off reply or hot take
7launch-coordinatorFull GTM sequence, readiness gates, run-of-showDelegate for anything bigger than a single feature announcement
8email-marketerLifecycle sequences, onboarding drips, subject linesDelegate for anything sent to more than a handful of users
9product-marketerChangelog entry, feature announcement, in-app blurbDelegate after every ship; DIY if the feature is trivial
10reddit-strategistSubreddit map, participation playbook, AMA prepDelegate before posting anywhere; it also flags astroturfing risk
11market-researcherJTBD/Mom Test synthesis, ICP, pain-point mapDelegate before positioning work; DIY if you've already run the interviews yourself
12customer-marketerCase studies from real wins, before/after narrativeDelegate once you have one approved customer quote; never invents quotes
13competitive-analystCompetitive matrices, battlecards, win-loss synthesisDelegate before a pricing change or new competitor entering your space
14brand-voiceTone matrix, terminology glossary, copy standardsDelegate once, early — then let other agents consume it
15pricing-editorPricing page packaging, tier names, pricing FAQDelegate for packaging; never let it touch live Stripe prices without explicit confirmation
16sales-closer15-person ICP list, first DM/email, day-3 follow-upDelegate for structured outbound; DIY for a single warm intro
17support-engineerRepro steps, minimal fix, customer-ready replyDelegate for any ticket without a stack trace already in hand
18video-producerDemo scripts, Loom outlines, shot lists, YouTube titlesDelegate before recording; never invents view counts or testimonials
19analytics-engineerEvent taxonomy, tracking plan, metrics martsDelegate when email-marketer or growth-strategist need new events instrumented
20help-writerHelp center articles, how-tos, troubleshooting guidesDelegate once a feature stabilizes; never invents UI that doesn't exist

1. seo-specialist

Owns keyword strategy, pillar/cluster architecture, page briefs, on-page checklists, and JSON-LD schema specs. It hands implementation specs to content-marketer and react-specialist — it does not write the full article itself. Trigger: new page, meta description, title tag, sitemap, or rankings work.

2. growth-strategist

Runs AARRR funnel diagnosis and maps growth loops, then prioritizes channels and experiments. Produces an experiment backlog consumed by copywriter, launch-coordinator, and positioning-strategist. It does not write code or implement tracking — that's analytics-engineer.

3. copywriter

Writes conversion copy — landing pages, headlines, CTAs, ad copy — using awareness-stage and CRO discipline, working from positioning-strategist and brand-voice inputs. Does not write React components or long-form SEO articles.

4. positioning-strategist

Runs Dunford-style positioning: category, competitive alternatives, differentiators, message house — from shipped capabilities only, never aspirational features. Runs on a stronger model than the rest of the team because the job is judgment, not drafting.

5. content-marketer

Plans SEO-aligned editorial calendars, pillar/cluster outlines, and lead magnets, then drafts the long-form pieces, coordinating with seo-specialist on briefs. Does not own homepage conversion copy or social cadence.

6. social-media-strategist

Plans LinkedIn-first B2B social strategy — 30-day calendars, hook libraries, post drafts, launch waves — applying brand-voice's tone row. Does not own Reddit strategy (reddit-strategist) or long-form blog (content-marketer).

7. launch-coordinator

Sequences a full go-to-market launch across positioning, content, email, SEO, social, and Reddit with owners, dependencies, and readiness gates. Assembles briefs for each specialist but does not write the channel copy itself.

8. email-marketer

Designs lifecycle email sequences, onboarding drips, and send logic with CAN-SPAM/GDPR compliance built in. Specifies the analytics events analytics-engineer implements. Does not build in-app onboarding UI or write cold outbound.

9. product-marketer

Turns shipped features into customer-language narratives — changelog entry, announcement email, social post, in-app blurb. Does not run multi-channel launches (launch-coordinator) or write 1:1 sales outreach (sales-closer).

10. reddit-strategist

Plans authentic, rules-compliant Reddit marketing — subreddit maps, participation playbooks, AMA prep, post drafts — and flags astroturfing or self-promotion risk before anything gets posted. Does not own LinkedIn strategy or primary customer interviews.

11. market-researcher

Runs JTBD and Mom Test discovery, synthesizes interviews, and defines the ICP and pain-point map. Feeds positioning-strategist and growth-strategist. Does not write marketing copy or run competitive teardowns.

12. customer-marketer

Turns one real customer win into a case study with a before/after narrative and approved quotes only. Never invents quotes or metrics. Does not own cold outreach (sales-closer) or feature changelogs (product-marketer).

13. competitive-analyst

Builds competitive matrices, battlecards, and win-loss synthesis from public sources and user-supplied notes. Produces date-stamped comparisons for positioning-strategist and copywriter. Does not write the positioning statement itself.

14. brand-voice

Defines brand voice, tone matrices, and terminology glossaries so every marketing artifact sounds consistent. Hands voice rules to copywriter, content-marketer, email-marketer, and social-media-strategist.

15. pricing-editor

Rewrites pricing page packaging, tier names, limits, and feature matrices without touching live Stripe prices unless the user explicitly confirms. Money-path changes go to billing-engineer.

16. sales-closer

Runs focused outbound — a 15-person ICP list, first DM or email, day-3 follow-up — without spam blasts. Does not write landing page copy (copywriter) or lifecycle nurture drips (email-marketer).

17. support-engineer

Turns pasted user complaints into repro steps, a minimal fix, and a customer-ready reply. Not a substitute for debugger when a stack trace and local failure already exist.

18. video-producer

Writes demo scripts, Loom outlines, shot lists, and YouTube titles/descriptions from real product capabilities. Never invents view counts or testimonials.

19. analytics-engineer

Owns product analytics — PostHog/Mixpanel tracking plans, event taxonomies, instrumentation, metrics marts. Not marketing SEO or legal compliance docs.

20. help-writer

Writes user-facing help center articles, how-tos, and troubleshooting guides from tickets or feature workflows. Never invents UI that does not exist.

How team-builder-style orchestration actually works in Claude Code

None of this requires a custom framework. Claude Code already has the primitives:

  • /agents lists every subagent definition available in the repo (the files under .claude/agents/) and lets you inspect or create one from the CLI.
  • Each agent file is a markdown file with YAML frontmatter — name, description, tools, model — the same shape shown in the roles above. The description field is what Claude Code's router matches against your request, which is why vague descriptions cause misrouting.
  • Parallel subagents run when a task genuinely splits into independent pieces — e.g., competitive-analyst researching alternatives while market-researcher synthesizes interviews — each in its own context, reporting back to the main thread rather than polluting it with intermediate research.
  • The handoff pattern that makes this a "team" and not a pile of prompts is files, not chat: growth-strategist writes an experiment backlog to disk, launch-coordinator reads it. That's what lets 20 roles stay coherent without one giant context window holding everything.

This is the same mechanism covered in more depth in our post on Agent Teams vs. subagents — marketing agents and engineering agents use the identical routing and handoff model, just with different roles.

Hand-assembling from marketplaces vs. installing the team

Building this yourself from a skills marketplace means picking 20 individual agent definitions, checking each one's tool grants aren't overlapping, writing the "does not" boundaries yourself, and making sure the handoff files (positioning.md, pricing.md, whats-new.md) actually land in predictable paths — then repeating that audit every time you add a 21st agent.

AgenticKit installs all 34 agents — the 20 marketing/growth roles above plus the engineering roles covered in our subagent roundup — in one command, with the tool scoping, "does not" boundaries, and file-path conventions already wired. You can see the full roster on the agent catalog page before installing anything.

FAQ

Do I need all 20 marketing agents on day one? No. Start with the five in the table above — seo-specialist, growth-strategist, content-marketer, email-marketer, and support-engineer — and add the rest as the workload actually calls for them (customer-marketer once you have a real win, sales-closer once you're ready for structured outbound).

What's the difference between an "agent team" and just prompting ChatGPT with different personas? A persona swap in one chat thread shares context and has no enforced boundary — the model can and will blend positioning advice into copywriting advice. A real agent team gives each role its own tool grant, its own "does not" line, and a file-based handoff, so positioning-strategist can't accidentally write landing page copy and copywriter can't redefine the category.

Can these agents work in Cursor instead of Claude Code? Yes — AgenticKit ships the same 34 roles as Cursor agent definitions (.cursor/agents/) alongside the Claude Code versions, though Claude Code's /agents routing and Agent Teams are the primary target.

Which agent should write my next blog post? content-marketer drafts it, seo-specialist writes the brief and schema spec first, and if it's going out as part of a bigger push, launch-coordinator sequences it against email and social. For a single reactive post, content-marketer alone is enough.

How is this different from a marketplace listing of individual marketing skills? A marketplace listing gives you the parts; it doesn't give you the ownership boundaries or handoff files that make 20 roles behave like one team instead of 20 competing opinions. That wiring is the actual product.

Sources

Written by Rakshit Yadav. Published 2026-10-31, last updated 2026-10-31.

Related Tools & Agents

🛠️ Free Tool: agent-chain🛠️ Free Tool: kit-picker🤖 Agent: seo-specialist🤖 Agent: growth-strategist🤖 Agent: content-marketer🤖 Agent: copywriter🤖 Agent: positioning-strategist🤖 Agent: email-marketer🤖 Agent: support-engineer⚡ Command: /seo-audit⚡ Command: /shipSkill: on-page-seoSkill: aarrr-diagnosticSkill: dunford-positioningSkill: pillar-cluster

See which marketing agent would actually run

Type a marketing task and watch the routing before you add a 21st agent.

Open Agent Chain →