---
title: "SEO Skill for AI agents and local SEO audits | Ian Nuttall"
description: "A local SEO CLI, MCP server, and agent skill for technical audits, Search Console analysis, competitor research, and evidence-backed fixes."
canonical: "https://ian.is/seo-skill"
language: "en"
---

# Run SEO audits your agent can actually act on

One local command turns crawl data, Search Console, Analytics, and keyword research into clear work with evidence behind it.

[Open SEO Skill](https://seoskill.dev/)

Node 22+ · Apache 2.0

[Star on GitHub](https://github.com/iannuttall/seo)

87 stars · Open source

## What an SEO audit should show before you change the site

### Start with a real crawl

Check metadata, links, canonicals, indexability, structured data, performance, mobile, and security from the site itself.

### Use the search data you own

Connect Search Console and Google Analytics so priorities come from queries, landing pages, and traffic rather than guesses.

### Give agents structured evidence

The CLI, MCP server, JSON output, and packaged skill all use the same report engine and stable finding IDs.

### Check whether the fix worked

Save crawl results, compare matched Search Console windows, and keep the verification step beside every recommendation.

### Reports stay inspectable

Observed data, derived findings, and recommendations remain separate so you can see how each conclusion was reached.

### No invented all-clears

Partial, capped, or missing data is labelled. A source that could not be checked is never reported as a zero.

## Local-first SEO work without another dashboard

Research is optional

DataForSEO, Semrush, and Ahrefs can add market estimates when they would change a decision. The main audit works without them.

Agents discover tools as needed

One short skill routes an agent to the right report instead of loading a long catalog into every context window.

Profiles and tokens stay local

Project profiles, connected-service tokens, reports, and caches live on your machine, not in someone else's dashboard.

Limits travel with the result

When a source is rate-limited or partially available, the report records that beside the data it did return.

## Common questions about SEO Skill

### Do I need Search Console to start?

No. `seo report --url` starts with a local technical crawl. Connect Search Console when you want query, position, and landing-page evidence.

### Can I use it without an AI agent?

Yes. The normal CLI path gives people the same reports. MCP and JSON output expose that engine to agents and scripts.

### Where does the data go?

Project profiles, tokens, reports, and caches stay on your machine. The tool only calls the services you connect and records their limits with the result.

### What is a technical SEO audit?

A technical SEO audit checks whether search engines can crawl, understand, index, and serve the useful pages on a site. SEO Skill combines those checks with search data so the report can rank what matters.

### Can an AI agent run the SEO audit?

Yes. The packaged skill, MCP server, CLI, and JSON output expose the same report engine. An agent gets stable finding IDs and the evidence behind each recommendation.

### Does it replace Ahrefs or Semrush?

No. SEO Skill can use third-party keyword and competitor data when it changes a decision, but its main job is combining your crawl, Search Console, Analytics, and local evidence.

### How do I verify an SEO fix?

Run the relevant report again and compare matching time windows where the source supports it. The report keeps observed data separate from conclusions so the before and after remain inspectable.

## Audit the site, rank the work, and give your agent evidence it can use.