SEOAgent automates the repetitive half of on-page SEO — titles, meta descriptions, schema, internal links, alt text — as real edits in your repo. Every change arrives as a diff you approve, so automation never publishes behind your back.
Your first SEO pull request in about 5 minutes. You pick the repo, and you merge every change.
Pick one above and we take you to the right setup.
The work is genuinely repetitive. The tools that automate it usually break something else.
Most automation rewrites titles and meta descriptions from the page text alone, with no idea which queries the page already ranks for. You get changes that read well and rank worse.
Unattended optimization edits live pages before anyone reads the change. By the time you notice a wrong claim or a mangled title, it is already indexed.
Dashboard tools change the rendered page, not the source. The next deploy overwrites the work, so the same fixes get reapplied forever and drift from what is in git.
Automate the mechanical work. Keep the judgement, and keep the approval step.
The Skill runs inside Claude Code, Cursor, or Codex and edits the actual files your build reads — frontmatter, head templates, JSON-LD components. The change survives your next deploy because it is in git.
Orphan pages and thin internal linking are the highest-leverage automated fix on most sites. SEOAgent maps what links to what, then proposes contextual links from the pages that actually carry authority.
The optional cloud layer reads your Google Search Console performance per page and prioritises by impressions, CTR, and position — so optimization targets the queries a page already surfaces for instead of the ones a tool assumed.
Every optimization arrives as a change you read, edit, or reject
Edits land in source files, not in a layer bolted onto the rendered page
Search Console data decides what to optimize first, not a generic checklist
It is software that performs repetitive on-page SEO work for you — writing or rewriting meta titles and descriptions, adding schema markup, generating image alt text, fixing heading structure, and proposing internal links. It is also called automatic SEO optimization, automatic search engine optimization, or SEO optimization automation; the terms are used interchangeably. The useful versions do this against evidence about how the page currently performs. The risky versions apply changes to live pages with no review.
In SEOAgent the changes are edits to files in your repo. Typical examples: the `title` and `description` in a blog post's frontmatter or a Next.js `metadata` export, a JSON-LD block in a layout component, `alt` attributes on images, a missing canonical tag, and contextual links added from one page to another. Each edit shows up in `git diff` before anything ships, so you can see exactly which lines the agent touched.
You run the Skill inside your own coding agent (Claude Code, Cursor, or Codex). It audits the repo, lists the pages with missing or duplicated metadata, thin internal linking, or absent schema, and edits the source files using your agent's own model. You read the diff, change what you disagree with, and merge through your normal pipeline. With the cloud layer connected, Google Search Console data decides which pages are queued first.
It is software that runs on your machine and inside your repo, not a managed service where an agency or a hosted tool edits your site for you. There is no second AI subscription and no per-credit metering: the work runs on the model you already pay for in your coding agent. The trade-off is that it needs a coding agent and a repo you can edit; it is not a dashboard-only workflow.
No. A manual action is a penalty applied by a human reviewer at Google, and it is cleared only by fixing the underlying problem and submitting a reconsideration request in Search Console. SEOAgent can help with the code-side work, such as finding thin or duplicated pages and removing spammy markup, but it does not file the request or guarantee the outcome.
A bare rewrite works from the page text alone. Automated optimization that uses Search Console starts from the queries a page already appears for, such as its impressions, click-through rate, and average position, and only then proposes a change. That matters most for pages sitting deep in the results: a new title rarely helps a page at position 40 or lower, so the better edit there is stronger content and internal links for the queries it already surfaces for.
Automate the mechanical and repetitive: alt text, schema blocks, canonical tags, internal-link discovery, and finding pages whose metadata is missing or duplicated. Keep humans on anything that makes a claim — product facts, pricing, positioning, and the decision about which topics are worth targeting at all.
No. SEOAgent proposes changes as diffs in your repository. You review, edit, and merge them, and they deploy through your normal pipeline. There is no mode where it edits your live site without an approved commit.
A plugin or dashboard usually modifies the rendered output or stores overrides in its own database, so your source of truth and your SEO drift apart. SEOAgent edits the source files themselves, which means the change is versioned, reviewable, revertible, and survives redeploys.
No. The Skill and the local CLI are free, need no account, and handle auditing plus the on-page edits. The $79 per site·month cloud layer adds Google Search Console analysis, competitor and keyword research, and evidence-backed prioritisation — it tells you which pages to optimize first.
Start by working out which of the three categories you are looking at — reporting, autoblogging, or execution with approval — because they are sold under one label and are not interchangeable. The buyer-side questions (where it writes, what it decides from, whether it can publish without you) are covered on the SEO automation software page at /seo-automation-software.
It can, if it rewrites pages that are already working or publishes unreviewed changes at scale. That is the main argument for a review step. Optimizing a page that ranks at position 8 needs more care than filling in metadata on a page with none — which is why prioritisation from real performance data matters more than the volume of changes made.
SEOAgent for GitHub
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.