A topical map is the cheapest SEO asset you can build and the easiest one to get wrong. Claude Code can draft one in a few minutes. The hard part is making the map reflect real demand, real SERPs and the site you actually have.
Start here: for SERP-based clustering, use the free seo-cluster skill in AgriciDaniel/claude-seo. For an entity-first map with a 90-day calendar, use the free Kai Topical Map skill. Either way, connect a keyword data MCP server (DataForSEO, Semrush or Ahrefs) before you trust the volumes. Paid packs are optional.
This guide covers the Claude Code skills worth installing for topical maps, both free and paid, what each one does, and how to install it. It also covers the parts a skill cannot do for you: deciding what belongs on the map, turning it into a publishing plan, and knowing when a map with no data behind it is just a guess in a spreadsheet.
Every repo below was checked on the GitHub API on 1 November 2026. Every price was checked on the live product page the same day.
What a topical map skill does in Claude Code
Short answer: a topical map skill is a SKILL.md file (or a plugin that bundles several) that teaches Claude Code a repeatable method for turning a seed topic into a structured plan: pillar pages, cluster articles, the entities each page must cover, internal links between them, and a publishing order. You call it by name, Claude follows its steps, and you get the same shape of output every time.
If you are new to skills, the basics are in what Claude Code skills are. The short version:
- A skill is a folder containing a
SKILL.mdfile with instructions and optional scripts or reference files. - Personal skills live in
~/.claude/skills/. Project skills live in.claude/skills/inside a repo. - Some skills ship as plugins installed through
/pluginmarketplaces. Others install withnpx skills add owner/repoor a plaingit clone. - Claude Code loads the skill when your request matches its description, or when you invoke it directly.
Why a skill beats a one-off prompt
A prompt like “make me a topical map for running shoes” gets you a list of article ideas. A good skill forces structure the prompt skips:
| Step | One-off prompt | Well-built skill |
|---|---|---|
| Seed research | Usually none | Web search, SERP checks, or keyword data via MCP |
| Clustering | Grouped by vibe | Grouped by intent or SERP overlap |
| Hierarchy | Flat list | Pillar, cluster and supporting pages |
| Entities | Rarely named | Entity list per node |
| Internal links | Missing | Link map from cluster to pillar and across siblings |
| Output | Chat text | Files you can version, edit and re-run |
That last row matters more than it looks. A topical map you can store in a repo, diff and update is a planning document. One that lives in a chat window is a brainstorm.
Takeaway: the skill is the method. The data you feed it decides whether the method produces anything useful.
The best Claude Code skills for topical maps in 2026
Short answer: for most people, start with the free seo-cluster skill in AgriciDaniel/claude-seo if you want SERP-based clustering, or the free Kai Topical Map skill if you want an entity-first map with a 90-day calendar. Add a keyword data MCP server (DataForSEO, Semrush or Ahrefs) once you need real volumes. Paid packs are optional.
Here is the shortlist. Stars and last-push dates come from the GitHub API on 2026-11-01.
1. claude-seo: seo-cluster and seo-plan (free)
- Source: github.com/AgriciDaniel/claude-seo
- Price: Free, MIT license
- GitHub: 25,381 stars, last push 2026-11-01
- What it does: the
seo-clusterskill clusters keywords by SERP overlap and designs a hub-and-spoke content architecture with internal links. Its own description says/seo cluster executethen writes the posts with the companion claude-blog skill if you have it installed, or produces content briefs if you do not. The repo also includesseo-plan,seo-content-briefandseo-dataforseo, so you can wire clustering to live data. - Install: follow the repo README. It supports a plugin install and a script install; pick one, not both.
This is the most complete free option for people who already think in SERPs. Clustering by SERP overlap (do the same URLs rank for both keywords?) is a sounder grouping rule than clustering by word similarity, because it reflects how Google already groups the queries.
2. Kai Topical Map (free)
- Source: github.com/cgallic/kai-cmo-harness, skill at
harness/skills/kai-topical-map/. Directory listing on claudskills.com - Price: Free, MIT license. The ClaudSkills directory sells an optional Pro tier for one-click installs ($9/month, $79/year or $149 one-time per its page); you do not need it to use the skill.
- GitHub: 60 stars, last push 2026-11-01
- What it does: builds what it calls an “AEO-first” topical map: entity clusters, query fan-out coverage, information gain scoring, a schema blueprint and a 90-day publishing calendar.
- Install:
mkdir -p ~/.claude/skills/kai-topical-map
curl -L https://claudskills.com/skills/kai-topical-map/SKILL.md \
-o ~/.claude/skills/kai-topical-map/SKILL.mdStart a new Claude Code session and it is discovered automatically. The repo is young and small, so read the SKILL.md before you rely on it. That advice applies to every skill on this list, but it matters most for low-star repos.
3. Hawk Academy Topical Authority Map (free)
- Source: Hawk Academy topical authority map guide, from StudioHawk’s education arm
- Price: Free, per the page
- What it does: the guide describes a skill that builds a four-cluster architecture (Core, AOR, Linking and Buyer Journey) and a 12-week publishing order from a seed keyword.
- Install: the guide links to a download page at
/claude-seo-skills/topical-authority-mapon the same site. The guide does not list install steps, so expect to drop the folder into~/.claude/skills/yourself.
Useful if you want an agency’s opinionated structure rather than a blank method.
4. marketingskills: site-architecture and content-strategy (free)
- Source: github.com/coreyhaines31/marketingskills
- Price: Free
- GitHub: 57,803 stars, last push 2026-10-31
- What it does: there is no skill named “topical map” here, but
site-architecture,content-strategy,programmatic-seoandai-seocover the same ground from a marketer’s angle: page hierarchy, content pillars and what to build first. - Install:
npx skills add coreyhaines31/marketingskills, or clone the repo into your skills folder.
Pick this if your map has to serve product marketing goals, not only rankings. More on that pack in Claude Code skills for marketers.
5. SEO Machine (free)
- Source: github.com/TheCraigHewitt/seomachine
- Price: Free
- GitHub: 7,565 stars, last push 2026-10-23
- What it does: a full Claude Code workspace for content operations. The repo root includes a
research_topic_clusters.pyscript alongside competitor-gap and SERP-analysis scripts, plus a DataForSEO test file. It is a working environment rather than a single skill. - Install: clone the repo and follow its QUICK-START.md.
6. Topical Authority Mapper (free)
- Source: listed on claudskills.com and on skills.sh from the nikiandr/goose-skills repo
- Price: Free
- GitHub: 1 star, last push 2026-03-27
- What it does: maps hub pages, spoke articles, supporting content and internal linking. The skills.sh version can mine Reddit for questions that become article topics.
- Install: copy the SKILL.md into
~/.claude/skills/topical-authority-mapper/.
Treat this as a reference to read and adapt. It has not been updated since March.
7. DataForSEO MCP server (paid usage, free server)
- Source: github.com/dataforseo/mcp-server-typescript
- Price: the server is free. Data is pay-as-you-go: $50 minimum top-up, $1 trial credit for new accounts, SERP API from $0.0006 per SERP, per dataforseo.com/pricing
- GitHub: 258 stars, last push 2026-11-01
- What it does: gives Claude Code live keyword volumes, SERPs and competitor data. This is what turns a guessed map into a measured one.
- Install: add it with
claude mcp addusing the command in the repo README and your API credentials.
8. Semrush MCP (paid)
- Source: semrush.com/mcp and Semrush developer docs
- Price: requires a Semrush One, SEO Classic or Trends API subscription, per Semrush’s MCP page. Plan prices change, so check the current pricing page rather than trusting a number in a blog post (including this one).
- What it does: keyword, domain and competitor data inside Claude Code, drawing from your plan’s API units.
- Install: the developer docs show
claude mcp add semrush https://mcp.semrush.com/v1/mcp -t http, then authenticate through/mcp.
Ahrefs also publishes an MCP server repo (100 stars, last push 2026-02-19). It needs an Ahrefs plan with API access; check Ahrefs for current terms. For Search Console data, a GSC MCP server lets the map include queries you already get impressions for.
9. Inflectual Master Marketing Skills Bundle (paid)
- Source: inflectual.gumroad.com
- Price: $167 one-time. Note that the product title also says “167 Claude AI Skills”. The price was confirmed separately from the page’s offer data (16,700 cents) and its own “$167 = ~$1 per skill” line, so the number match is a deliberate pricing choice, not a misread.
- What it does: a large bundle of marketing skills. The sales page does not itemise a dedicated topical map skill, so confirm what you get before buying if that is your only reason to buy.
- Install: download from Gumroad and copy the folders into
~/.claude/skills/.
No fake countdowns or “spots left” claims appeared on the page when checked.
What about Contra and other paid packs? Searches for a paid topical map skill on Contra returned nothing specific enough to verify. Directory aggregators like ComposioHQ/awesome-claude-skills (78,501 stars), alirezarezvani/claude-skills (29,110 stars) and sickn33/agentic-awesome-skills (48,967 stars) are good places to look for new options. Anthropic’s official plugins directory (45,680 stars) is the place to check for vetted plugins, though it had no dedicated topical map plugin at the time of writing.
Free vs paid: which topical map setup to pick
Short answer: the skills themselves are almost all free. What you pay for is data. A free skill plus a pay-as-you-go data source beats a paid skill pack with no data every time.
| Option | Type | Price | Data built in? | Best for |
|---|---|---|---|---|
claude-seo seo-cluster | Skill | Free | Optional via DataForSEO skill | SERP-overlap clustering |
| Kai Topical Map | Skill | Free | No, uses web search | Entity-first maps, 90-day calendar |
| Hawk Academy Topical Authority Map | Skill | Free | No | Agency-style 4-cluster structure |
| marketingskills | Skill pack | Free | No | Maps tied to product marketing |
| SEO Machine | Workspace | Free | Via DataForSEO | Full content ops in a repo |
| Topical Authority Mapper | Skill | Free | No | Reading and adapting |
| DataForSEO MCP | Data | $50 min top-up, $1 trial credit | Yes | Cheap, metered keyword data |
| Semrush MCP | Data | Semrush One, SEO Classic or Trends API plan | Yes | Teams already on Semrush |
| Ahrefs MCP | Data | Ahrefs plan with API access, pricing not public here | Yes | Teams already on Ahrefs |
| Inflectual bundle | Skill pack | $167 one-time | No | Buyers who want many marketing skills at once |
A sensible stack for a solo operator: one free mapping skill, DataForSEO on pay-as-you-go, and Google Search Console for your own queries. Install steps for each are in how to install Claude Code skills.
Pillar and cluster structure: what the skill should produce
Short answer: a usable map has three levels. Pillars cover a broad topic and target a head term. Clusters each answer one intent in depth and link up to their pillar. Supporting pages handle narrow questions and link to the cluster they serve. A skill that only gives you a flat list of titles has not built a map.
The three levels
- Pillar. One page per major topic. It defines the topic, summarises each subtopic and links to every cluster underneath. Example: “Claude Code skills for SEO”.
- Cluster. One page per distinct search intent. Example: “Claude Code skills for keyword clustering”. It should rank on its own merits, not exist to feed the pillar.
- Supporting. Narrow pages that answer one question, such as “how to install a skill from a GitHub repo”. These pick up long-tail traffic and pass context up the chain.
How to tell if the clustering is right
Run this check on any map a skill produces:
- One intent per page. If two planned pages would rank for the same SERP, merge them. This is the main value of SERP-overlap clustering, covered in depth in Claude Code skills for keyword clustering.
- No orphan clusters. Every cluster has a parent pillar.
- No giant pillars. If a pillar has 40 clusters, it is really two or three pillars.
- Links go both ways. Cluster to pillar, pillar to cluster, and sibling to sibling where the reader would genuinely want the next page.
The internal link map is where most AI-generated maps are weakest. They list links, but the anchors are generic and nothing checks whether the target page exists. Pair the map with a linking skill from Claude Code skills for internal linking once pages are live.
Entity coverage: the part most maps skip
Short answer: entity coverage means listing the people, products, concepts and attributes each page must mention for the topic to be complete. Search engines and AI answer engines both build understanding from entities. A map that names pages but not entities leaves the writer to guess what “complete” means.
What an entity list looks like
For a cluster page on “Claude Code skills for topical maps”, the entity list would include:
| Entity type | Examples |
|---|---|
| Product | Claude Code, SKILL.md, MCP |
| Tools | DataForSEO, Semrush, Ahrefs, Google Search Console |
| Concepts | topical authority, pillar page, content cluster, search intent, SERP overlap |
| Actions | install, cluster, map, publish, refresh |
| Places | GitHub, skill directories, Gumroad |
How to get entities that are real, not invented
Ask the skill to pull entities from the pages that already rank, not from its own memory. A practical prompt pattern inside Claude Code:
- Fetch the top results for the cluster keyword (with WebSearch or a SERP MCP).
- Extract the named concepts and tools that appear in most of them.
- Mark which ones your planned page covers and which it misses.
- Add the gaps to the brief.
This is the same logic the Kai skill describes as “query fan-out coverage”: AI search systems often break one question into several sub-queries, so a page that answers the neighbours of the main question has more chances to be cited. For more on that side, see Claude Code skills for GEO.
Takeaway: pages without entity lists get written from the writer’s memory. Pages with entity lists get written against the SERP.
Turning a topical map into a publishing plan
Short answer: sort the map by three things (business value, ranking difficulty and dependency), then publish pillars and the clusters that support them in batches, so each batch forms a complete, interlinked unit. A map is a list of what to write. A publishing plan says what to write first and why.
A simple scoring sheet
Score each planned page from 1 to 3 on each column:
| Page | Business value | Ease (inverse of difficulty) | Has data? | Total | Batch |
|---|---|---|---|---|---|
| Pillar: Claude Code skills for SEO | 3 | 1 | 3 | 7 | 1 |
| Cluster: keyword clustering | 3 | 2 | 3 | 8 | 1 |
| Cluster: topical maps | 3 | 2 | 2 | 7 | 1 |
| Supporting: install from GitHub | 1 | 3 | 2 | 6 | 2 |
The numbers here are illustrative, not measured. Your own sheet should use volumes and difficulty from your data source.
Batch, don’t drip
Publish a pillar with at least a few of its clusters together. A lone pillar with no clusters has nothing to link to, and a lone cluster with no pillar has nowhere to send authority. Batches make internal links real on day one.
The weekly loop
- Pick the next batch from the sheet.
- Generate a brief for each page (Claude Code skills for content briefs).
- Write and edit (Claude Code skills for content writers).
- Add schema and internal links.
- After a few weeks, check Search Console and update the map: merge pages that cannibalise, add pages for queries you rank for by accident.
- Refresh older pages on a schedule (Claude Code skills for content refresh).
Kai’s 90-day calendar and Hawk Academy’s 12-week order are both versions of this. Use theirs as a starting draft and reorder it with your own scores.
The limits of a topical map with no data behind it
Short answer: a map built only from a language model’s knowledge has no volumes, no difficulty and no view of the live SERP. It can describe a topic well and still send you to write pages nobody searches for, or pages that duplicate one another in Google’s eyes. Treat a data-free map as a hypothesis.
What goes wrong without data
- Phantom demand. The model proposes a cluster because it sounds logical. Nobody searches for it.
- Hidden cannibalisation. Two pages look different to the model but share the same SERP.
- Stale entities. Tools and product names from the model’s training data may be out of date. Claude Code skills move fast; several repos above were pushed to on the day this was written.
- No view of your site. Without crawl or Search Console data, the map ignores pages you already have, so you plan rewrites as new posts.
- No competitor reality. Some clusters are owned by sites you cannot outrank this year. Only SERP data tells you which.
The minimum data to add
| Data | Source | What it fixes |
|---|---|---|
| Keyword volume and difficulty | DataForSEO, Semrush or Ahrefs MCP | Phantom demand, prioritisation |
| Live SERPs | SERP API or WebSearch | Cannibalisation, entity gaps |
| Your queries and pages | Google Search Console MCP | Existing coverage, quick wins |
| Your site crawl | An audit skill (Claude Code skills for SEO audits) | Duplicate and orphan pages |
| Forum questions | Reddit, Quora, PAA | Real phrasing and objections |
You do not need all five on day one. Volume plus live SERPs removes the biggest errors for a small cost. Keyword sourcing is covered in Claude Code skills for keyword research.
What a skill still cannot decide
Even with perfect data, a skill does not know which topics your business can credibly own, which products you sell, or what you have real experience to write about. Topical authority comes from covering a topic well and consistently, and that depends on having something true to say. That judgement stays with you.
Common mistakes when using Claude Code for topical maps
Short answer: most failures come from trusting the first output, mapping too wide, and skipping the review step. The fixes are cheap.
- Accepting the first draft. Run the skill, then ask Claude to critique its own map against the SERP. The second pass usually merges duplicates.
- Mapping the whole industry. A 300-page map for a site with 12 posts is a wish list. Map one pillar deeply.
- Installing five overlapping skills. Two mapping skills with similar descriptions confuse Claude about which to load. Keep one.
- Not reading the SKILL.md. Skills can run scripts. Read what you install, especially from low-star repos.
- Skipping the link map. A map without planned internal links is a list of titles.
- Never updating it. The map should change after every Search Console review.
- Buying a pack for one skill. Check the contents list first.
If you want a mapping method tailored to your niche, building your own skill is not hard: how to build a Claude Code skill. For scale beyond hand-written pages, see Claude Code skills for programmatic SEO.
FAQ
What is a topical map in SEO?
A topical map is a structured plan of every page a site needs to cover a subject completely, organised into pillars, clusters and supporting pages, with the internal links between them.
Can Claude Code create a topical map?
Yes. With a skill such as claude-seo’s seo-cluster or Kai Topical Map, Claude Code can produce a structured map with clusters, entities and a publishing order. Its accuracy depends on the data you connect.
What is the best free Claude Code skill for topical maps?
For SERP-based clustering, seo-cluster in AgriciDaniel/claude-seo. For an entity-first map with a 90-day calendar, Kai Topical Map. Both are free and MIT licensed.
Do I need a paid SEO tool to build a topical map with Claude?
No, but you will get a guess rather than a plan. DataForSEO’s pay-as-you-go model ($50 minimum top-up) is the cheapest way to add real volumes.
How do I install a topical map skill?
Copy the skill folder into ~/.claude/skills/ (personal) or .claude/skills/ (project), or use npx skills add owner/repo or a /plugin marketplace if the author supports it. Start a new session so Claude discovers it.
What is the difference between a topical map and keyword clustering?
Keyword clustering groups keywords that one page can target. A topical map arranges those pages into a hierarchy with links and a publishing order. Clustering is one input to a map.
How many articles do you need for topical authority?
There is no fixed number, and no reliable source gives one. Cover every distinct intent in your topic that your audience searches for. For a narrow niche that might be a few dozen pages.
What is a pillar page?
A broad page that covers a whole topic at summary level and links to every cluster page underneath it.
How long does it take to see results from a topical map?
It varies by site authority and competition, and no honest source can promise a timeline. Review Search Console a few weeks after each batch and adjust.
Should a topical map include entities?
Yes. An entity list per page tells the writer what “complete” means and helps search and AI engines connect the page to the topic.
Can I use Semrush or Ahrefs data inside Claude Code?
Yes, through their MCP servers. Semrush says MCP needs a Semrush One, SEO Classic or Trends API subscription. Ahrefs requires a plan with API access.
Are paid Claude Code skill packs worth it for topical maps?
Rarely, for this one job. The strongest mapping skills are free. Paid money is better spent on keyword data.
Is it safe to install skills from GitHub?
Read the SKILL.md and any scripts before installing. Skills can run code with your permissions, so prefer known repos and check recent activity.
Can Claude Code build a topical map for an existing site?
Yes, if you give it the site’s data: a crawl or sitemap and Search Console queries. Without that, it plans as if the site were empty.
Does a topical map help with AI search and GEO?
It can. Covering the neighbouring questions of a topic gives AI answer engines more passages to cite. See the GEO guide linked above.
Sources and methodology
- GitHub star counts and last-push dates: GitHub REST API, queried 2026-11-01.
- Skill behaviour: each repo’s own SKILL.md or README, read on 2026-11-01. Descriptions here paraphrase the authors’ claims; they were not independently benchmarked.
- Prices: DataForSEO pricing, Semrush MCP, Inflectual on Gumroad, ClaudSkills Kai listing, all checked 2026-11-01.
- Skill directories searched: claudskills.com, skills.sh, GitHub awesome lists. Searches of Reddit and Contra did not surface a verifiable topical map product.
Your next step
Pick one pillar you can genuinely own. Install one free mapping skill, connect one data source, and generate a map for that pillar only. Score the pages, publish the first batch together, and revisit the map after your next Search Console review. For the wider toolkit, start with the full list of Claude Code skills for SEO.
