BeyondSEO released an open-source SEO skill for Claude Code and Codex that ships its own crawler and needs no paid API

BeyondSEO is a new open-source SEO toolkit that runs as a skill inside Claude Code, Codex, Claude Desktop and ChatGPT Work, and it ships its own Python crawler so the audit needs no third-party SEO API. The project's author, Tahir, first pushed it to GitHub on 13 September under an MIT licence, and the latest tagged release, v2.7.1, went up on 16 September. The repository sits at 90 stars, 24 forks and no open issues as of today.

The design is unusual for the space. Most SEO agent tools call a paid backend such as Semrush, Ahrefs or DataForSEO. BeyondSEO ships a native crawler that reads sitemaps and raw HTML, renders JavaScript with a local Chromium through Playwright when needed, and stores its findings in SQLite so a snapshot can be resumed. The README says the CLI itself makes no language-model calls and consumes no model tokens. The assistant's own model spend still applies when it explains findings or writes copy.

What it covers

BeyondSEO groups its work under six headings: inspect the website, explain what matters, build answer-ready content, verify reputation, make the plan actionable, and improve and review. That maps to technical SEO, Answer Engine Optimization, Generative Engine Optimization, entity and authority signals, backlink discovery and verification, and named-competitor comparison. The reputation module refuses to count a search snippet as a verified backlink, a distinction most free tools skip.

The v2.7.1 release notes list four changes. Competitor selection now checks the target business's services and customers before comparing, and separates direct competitors from international benchmarks. Backlink research verifies source pages through the crawler. Reports export as branded PDF and self-contained HTML with contents, evidence tables and a 30/60/90-day roadmap. The skill ZIP was cut from 211 to 196 files to sit under Claude's 200-file skill-upload limit.

How it installs

The distributable is a single skill bundle. On Claude web or desktop, an operator uploads the beyondseo-2.7.1-skill.zip release asset through Customize, Skills. On Claude Code and Codex, an included Python installer registers the skill. On ChatGPT Work the same ZIP goes in through Customize, Skills, Create. Six hosts are named in total, plus a straight Python interface for anyone who wants to use the crawler on its own.

The author notes on the release page that ChatGPT Work's own execution environment still hits browser-download and DNS failures during the crawler bootstrap. Users on those hosts can point the assistant at a separate approved runtime folder for the crawl step.

What it does not claim

The README refuses several claims a commercial SEO product would make: no hosted crawling account, no SEO-data subscription, no ranking guarantee, and no unmeasured speed comparison against another crawler. Automatic mode is called out as a heuristic, browser mode's scroll and wait behaviour is not called complete, and robots rules are respected by default. Conservative scoring is a stated design goal: unknown evidence stays unknown rather than being counted as a high authority claim.

For a working SEO consultant or an in-house content team, BeyondSEO takes the pieces normally scattered across a crawler subscription, a backlink dashboard, a schema tool and a briefing document, and puts them into one skill that the same coding agent already loads for the rest of the work. The traction is early, and any team looking at it should read the release notes for what the tool says it does not do as carefully as what it says it does.

Source

Primary source: beyondtahir/beyondseo on GitHub. Release notes: BeyondSEO 2.7.1.