# Claude Code Role Agents vs Cursor Custom Agents

> Looking for a Claude Code SEO agent, growth agent, support-triage agent, or QA agent? Each is a .claude/agents/<name>.md role with its own tools and model. What each one does, which skills it's wired to, and how installing it differs from Cursor's @-mentioned agent files.

Canonical: https://www.agentickit.co/compare/claude-code-role-agents-vs-cursor-custom-agents
Last updated: 2026-11-18

A "Claude Code SEO agent" or "Claude Code growth agent" isn't a separate product — it's a subagent file at .claude/agents/<name>.md with a scoped tool grant, a model choice, and a system prompt for that one job. AgenticKit ships one for each of those roles already: seo-specialist, growth-strategist, email-marketer for lifecycle, support-engineer for ticket triage, and qa-explorer for exploratory QA. Cursor's equivalent lives at .cursor/agents/<name>.md and is invoked by @-mentioning the role in your current chat instead of delegating to an isolated context window. Last updated 2026-11-01 by Rakshit Yadav.



## There's no separate "SEO agent" product — it's a role file
When someone searches for a "Claude Code SEO agent" or "Claude Code growth agent," there's no dedicated SKU to buy. What they're actually looking for is a [subagent](https://www.agentickit.co/compare/claude-code-skills-vs-subagents-vs-hooks): a markdown file at `.claude/agents/&lt;name&gt;.md` with frontmatter for `name`, `description`, a `tools` grant, and a `model`, plus a system prompt scoped to that one job. The main Claude Code session delegates to it either because you named it or because the task matches its description, and it runs in its own isolated context window before reporting back. AgenticKit's roster already has one role file for each of the five jobs people usually mean: [seo-specialist](https://www.agentickit.co/agents/seo-specialist) (SEO lead), [growth-strategist](https://www.agentickit.co/agents/growth-strategist) (growth PM), [email-marketer](https://www.agentickit.co/agents/email-marketer) (the closest fit to "lifecycle marketer" — onboarding drips, activation nudges, win-back), [support-engineer](https://www.agentickit.co/agents/support-engineer) (ticket triage), and [qa-explorer](https://www.agentickit.co/agents/qa-explorer) (exploratory QA review).

## What each role actually does and hands off
<table><thead><tr><th>Query</th><th>Agent</th><th>Does</th><th>Hands off to you</th></tr></thead><tbody><tr><td>SEO lead</td><td>[seo-specialist](https://www.agentickit.co/agents/seo-specialist)</td><td>Technical SEO audits, JSON-LD schema, keyword clustering, programmatic page architecture</td><td>Publishing decisions and any claim that needs a real source</td></tr><tr><td>Growth PM</td><td>[growth-strategist](https://www.agentickit.co/agents/growth-strategist)</td><td>AARRR funnel mapping, referral loop design, free-tool and activation experiments</td><td>Which experiments actually ship and get budget</td></tr><tr><td>Lifecycle marketer</td><td>[email-marketer](https://www.agentickit.co/agents/email-marketer)</td><td>Welcome drips, activation nudges, trial-expiry and win-back sequences</td><td>Sending the first real batch — it drafts, it doesn't hit send</td></tr><tr><td>Support triage</td><td>[support-engineer](https://www.agentickit.co/agents/support-engineer)</td><td>Turns a user complaint into a repro, a minimal fix, a regression test, and a plain-language reply</td><td>Judgment calls on refunds, escalations, or anything outside the repro'd bug</td></tr><tr><td>QA reviewer</td><td>[qa-explorer](https://www.agentickit.co/agents/qa-explorer)</td><td>Clicks through a flow like a new user, files a ranked bug list with severity and repro steps</td><td>Deciding which bugs block a release</td></tr></tbody></table>Each agent page lists the exact `tools` and `model` grant parsed from the real `.claude/agents/&lt;slug&gt;.md` file, plus the skills it's wired to read before it starts — see the agent pages linked above for the current values, since those can change between kit releases.

## How this differs from Cursor's version
Cursor's side of this is structurally similar but mechanically different, as covered in [Cursor Agents vs Commands](https://www.agentickit.co/compare/cursor-agents-vs-commands): a role file at `.cursor/agents/&lt;name&gt;.md` is still a brief — what the job is, what it reads, what it's not allowed to do — but you bring it into a job by @-mentioning it inside your current chat, not by delegating to a context window the main thread can't see. There's no equivalent of the clean-context isolation that makes a Claude Code subagent's scratch work disappear once it reports back. If you run the same five-role setup in Cursor, @-mention `seo-specialist` or `support-engineer` the same way you'd @-mention any teammate, and expect its intermediate steps to stay visible in the thread.

## Can you install just the one role you need?
Not at the single-agent level — AgenticKit installs at the kit level, not per role. `npx agentickit-cursor init` (or the Claude Code equivalent) writes the **Marketing Kit** or the **Engineering Kit** into your repo, and each kit brings its full set of role files with it: installing the Marketing Kit for the SEO or growth role also gives you email-marketer, content-marketer, and the rest of the marketing roster, not just the one you asked for. support-engineer and qa-explorer ship in the Engineering Kit. If you only ever plan to use one or two roles from a kit, that's fine — the unused role files cost nothing until you @-mention or delegate to them — but there's no narrower install than "the kit this role lives in."

## Example prompt and what comes back
For seo-specialist: `"You are seo-specialist. Audit /pricing for missing schema and thin metadata, then propose three programmatic page slugs we could add to lib/comparisons.ts."` A reasonable before/after: before, a pricing page with no JSON-LD and a generic meta description; after, the agent returns a short audit (what's missing, where), a proposed `Product` or `FAQPage` schema block, and three slug ideas with the exact query each would target — not code it has already committed, since on-page changes still go through review like any other agent output.



## FAQ

### Is there a dedicated Claude Code SEO agent I can install by itself?
There's a dedicated role file — seo-specialist — but it installs as part of the Marketing Kit, not as a standalone package. See the role-by-role table above for what it does and which skills it reads.

### What's the closest Claude Code agent to a "lifecycle marketer"?
email-marketer. AgenticKit doesn't have a role named lifecycle-marketer; email-marketer covers the same ground — welcome drips, activation nudges, trial-expiry reminders, and win-back sequences.

### Does a support-triage or QA agent write code directly?
support-engineer and qa-explorer both produce artifacts (a patch plus a regression test, or a ranked bug list) that a human or a separate review agent checks before anything ships — see each agent's page for its exact tool grant.

### How does a Claude Code role agent hand off to the rest of the 34-agent team?
Through a command or the main session's own delegation: a command like /ship or /launch can invoke tech-lead, then a specialist role, in sequence, and each subagent reports back before the next one starts — see Claude Code Agents vs Commands for how that chaining works.

## Related
- [The 34 AgenticKit agents](https://www.agentickit.co/agents)
- [seo-specialist agent](https://www.agentickit.co/agents/seo-specialist)
- [growth-strategist agent](https://www.agentickit.co/agents/growth-strategist)
- [email-marketer agent](https://www.agentickit.co/agents/email-marketer)
- [support-engineer agent](https://www.agentickit.co/agents/support-engineer)
- [qa-explorer agent](https://www.agentickit.co/agents/qa-explorer)
- [Claude Code Agents vs Commands](https://www.agentickit.co/compare/claude-code-agents-vs-commands)
- [Cursor Agents vs Commands](https://www.agentickit.co/compare/cursor-agents-vs-commands)
- [AgenticKit vs Cursor Background Agents](https://www.agentickit.co/compare/agentickit-vs-cursor-background-agents)
