Framework-agnostic SEO infrastructure for JavaScript applications. The engine
turns site configuration and route data into consistent metadata, JSON-LD,
robots.txt, llms.txt, sitemap entries, redirects, content analysis, audits,
and approval-safe SEO change proposals.
It is designed for applications that own their source code. Runtime output is
generated by your application; the engine does not require WordPress, a CMS,
Prisma, Redis, or a hosted GMB Ranker account.
Framework-agnostic SEO engine — metadata, schema, audits, safe changes, MCP contracts, and content analysis
The Next adapter also provides createLlmsRoute, createSitemapHandler, and
createNextMetadata/toNextMetadata. See the adapter README for complete
route examples.
What the core provides
Metadata templates: title, description, canonical, robots, Open Graph,
Twitter/X, locale, and alternate-language URLs.
Discovery files: robots.txt policies, AI crawler presets, llms.txt, and
sitemap contracts.
Content analysis: keyword parsing and density, readability, title and
description limits, headings, links, images, and deterministic SEO scores.
Technical helpers: redirect matching and graph inspection, URL normalization,
link counting, internal-link proposals, and broken-link classification.
Decision contracts: capability discovery, privacy-safe site manifests,
validated metadata/social/schema/redirect/link/robots changes, audit reports,
and change history interfaces.
Storage interfaces: application-owned adapters for settings, metadata,
redirects, robots content, sitemaps, and change history.
Decision engine and MCP boundary
The package does not directly mutate a customer website. It provides typed
evidence and change contracts that a host, CMS adapter, MCP server, or CI
pipeline can inspect, approve, execute, and verify.
Treat MCP or AI output as a proposal. The host application remains responsible
for authentication, authorization, approval, rate limits, secrets, persistence,
rollback, and deployment.
Capability status
SEO_ENGINE_CAPABILITIES is the public capability registry shared by the
engine, adapters, CLI, and MCP integrations. Each capability has a status:
stable: implemented and covered by the package contract;
beta: usable but still evolving;
planned: product roadmap or host integration, not a runtime promise.
Do not advertise a planned capability as implemented. CMS connectors,
Google Search Console/Analytics/Business Profile providers, and MCP transport
are host integrations rather than hidden dependencies of the core package.
Storage
The core is deliberately persistence-free. Implement the interfaces in
STORAGE.md using your database, filesystem, CMS, or service.
Never put credentials or provider tokens in the SEO manifest or public JSON-LD.
The CLI reports pass, warning, and fail findings. Warnings do not fail the
audit; missing titles and descriptions do. Use it in CI after your site is
deployed or against a preview URL.
Development
npm install
npm run build
npm test
npm run check:publish
check:publish builds and tests every workspace and performs a dry-run pack
for each publishable package. Releases are published by the GitHub Actions
workflow from version tags with npm provenance. Never commit an npm token or a
credential-bearing .npmrc.