What is the SEO Skill for Codex?
The SEOAgent Skill turns OpenAI Codex into an SEO engine for your codebase. Once installed, you can ask Codex to audit your site, fix technical issues, research keywords, and write landing pages and articles — and it does the work as real files in your repo. Nothing publishes on its own: every change comes back as a normal diff to approve before you commit. It is the opposite of an auto-blogging tool — it is SEO that lives where you already ship.
Three steps, about five minutes
- Connect GitHub. Pick the repo you work on in Codex. GitHub asks which repos SEOAgent may see, and that is the whole signup.
- Merge the setup pull request. It opens in about 5 minutes with what SEOAgent found on your site and its plan. Review it in Codex or on GitHub.
- Get one pull request every weekday. Fixes and articles, each one yours to edit and merge.
seoagent init --code, then “Logged in as” — the site is bound.How to use SEOAgent in Codex
Install the Skill globally
Run `npm install -g @seoagent-official/seoagent` once, then `seoagent init` in your project. It writes the Skill bundle to `.agents/skills/seoagent/` when your repo already uses `.agents/`, and to `.claude/skills/seoagent/` otherwise, then adds a SEOAgent section to the repo-root `AGENTS.md` pointing at that file. `AGENTS.md` is the file Codex reads, so it picks the Skill up without any config. `init` prints the exact path it used.
Ask Codex to audit your site
Tell Codex "audit my site." The Skill fetches your homepage, sitemap, and robots.txt, runs a full technical-SEO check, and reports the biggest issues with the exact fix for each — not a vague to-do list.
Approve the fixes
Codex proposes changes — metadata, schema, canonicals, internal links — as edits you review before committing. Apply what makes sense and skip the rest. Safe, reversible, shipped through your normal pipeline.
Write content in your repo
Ask it to research keywords and write articles or landing pages. It builds a topic cluster with internal linking and writes the files straight into your repo (or your CMS), so a reviewed commit puts them live.
Connect Search Console (optional)
Run `seoagent login` to connect Google Search Console. Now the Skill grounds its suggestions in your real traffic data — which queries sit on page two, which pages are slipping — instead of guessing.
Give it a cadence
SEO compounds on a schedule, not on one good afternoon. A weekday cron line — `0 9 * * 1-5 cd /path/to/repo && seoagent sync --silent` — keeps the queue of detected work fresh; then open Codex in the repo and say "process the inbox" to do the work. Everything it learns is written to `.seoagent/`, so the next session starts where the last one stopped instead of re-auditing from zero.
What SEOAgent does in Codex
Technical SEO audits
Crawls your pages and flags indexing, canonical, metadata, render, and internal-link issues — with the fix attached to each finding.
In-repo fixes
Applies safe, mechanical fixes (titles, meta descriptions, schema, alt text) directly in your source, and proposes the rest for review.
Keyword & competitor research
Pulls real volume and difficulty and finds the queries you can realistically win, so content targets terms that matter.
Content that ranks and gets cited
Writes landing pages and articles structured for both readers and AI answer engines, with automatic internal linking into a topic cluster.
Search Console grounding
Reads your real GSC performance and suggests evidence-backed updates, so the work compounds instead of going stale.
Human-in-the-loop, always
Every change is a diff you review before committing. No autopilot, no AI slop published behind your back.
Orphan and internal-link analysis
`seoagent internal-links` reads the repo, not a crawl, and reports which pages have no inbound internal links and which have exactly one — the pages Google discovers and then declines to index.
AI search readiness
Generates the machine-readable layer answer engines actually read: an `llms.txt` built from your real page inventory (`seoagent llms`) and an Open Knowledge Format bundle served at `/.well-known/okf/` (`seoagent okf`).
Frequently asked questions
How do I add SEO to Codex?
Install the SEOAgent Skill: run `npm install -g @seoagent-official/seoagent`, then `seoagent init` in your project. `init` adds a SEOAgent section to `AGENTS.md`, which is how Codex finds the Skill — then just ask it to "audit my site" to start.
Is the SEOAgent Skill for Codex free?
Yes. The Skill is free forever and runs locally. An optional cloud tier ($79 per site per month) adds Google Search Console analysis, competitor research, and cloud review of suggested updates.
Does it publish content automatically?
No. SEOAgent proposes every change as a diff you review and approve before committing. It writes files into your repo or CMS, but you decide what ships — it is intentionally not an auto-publishing blog spammer.
What can Codex actually do with the Skill?
Run technical SEO audits, fix metadata/schema/canonicals/internal links in your source, research keywords and competitors, and write optimized landing pages and articles into your codebase — grounded in your Search Console data once connected.
What is the best SEO skill for Codex?
SEOAgent is the highest-scoring SEO skill on the public SEO Skill Benchmark, and the only one built as a full harness: audits, fixes, keyword research, content, and Search Console grounding in one free install that never publishes without your approval.
Can Codex do SEO on its own?
Codex can edit any file, but out of the box it has no SEO playbook: no audit checklist, no keyword data, no Search Console access. The Skill supplies that missing harness, so "audit my site" produces a real technical-SEO report instead of generic advice.
Where does the SEO skill install for Codex?
`seoagent init` writes the Skill bundle to `.agents/skills/seoagent/` if your repo already uses `.agents/`, and to `.claude/skills/seoagent/` otherwise. It then appends a SEOAgent section to the repo-root `AGENTS.md` that points Codex at the installed `SKILL.md` — that section is the channel Codex reads. The files are tracked in git, so a teammate who clones the repo gets the same Skill without installing anything.
How does Codex know the SEO skill is there?
Through `AGENTS.md`. `seoagent init` adds a SEOAgent section telling Codex to read the installed `SKILL.md` at session start, run `seoagent sync`, and triage `.seoagent/inbox/`. The step is idempotent: if `AGENTS.md` already mentions seoagent — even a section you wrote yourself — `init` leaves the file untouched.
How do I update the Codex SEO skill?
Run `npm install -g @seoagent-official/seoagent@latest`, then `seoagent init` again in the project. Re-running `init` refreshes the Skill bundle in place and leaves everything under `.seoagent/` — your audits, keywords, briefs and roadmap — untouched.
Does it work with Claude Code and Cursor too?
Yes. The same Skill works in Claude Code and Cursor, plus a standalone CLI. See the SEO Skill for Claude Code and for Cursor.
See also
SEOAgent for GitHub
Your first SEO pull request in about 5 minutes.
Connect the repo your site lives in. SEOAgent finds what holds the site back in search and fixes it in pull requests. You review and merge every change.
- ✓14 days free
- ✓No card
- ✓$79/site/month after
- ✓What SEOAgent found, and the plan for your site
- ✓Meta descriptions for 14 pages
- ✓Sitemap and robots.txt
- ✓Organization schema
- 370+sites
- 1,100+articles written
- 3,500+SEO changes made