Metrifyr 2.0 is out. This release bundles three things that have been shipping over the past two weeks into a single coherent step up: a redesigned homepage and SEO surface, a full SEO playbook library served natively from the MCP server, and two new tools that fill real gaps in modern SEO and GEO work.

Here is what changed and why it matters.

1. SEO skills library, served natively

Metrifyr used to be tools-only. You connected it, you got ~70 Google Marketing tools. That is great raw material, but it is not a playbook. If you wanted structured SEO work — keyword research, a content refresh pass, a rank tracker sweep, a technical SEO check — you had to bring your own prompts or pull in an external skills library with npx, configure it, hope the connectors matched.

In 2.0, Metrifyr serves the full SEO skills library directly as MCP prompts. Sixteen playbooks, available the moment you connect:

research — keyword research, competitor analysis, content gap analysis, SERP analysis
optimise — on-page SEO auditor, technical SEO checker, content refresher, internal linking optimiser
monitor — rank tracker, backlink analyser, performance reporter, alert manager
cross-cutting — content quality auditor, domain authority auditor, entity optimiser, memory management

In Claude.ai, Cursor or any MCP-aware client, open the prompt picker and type @metrifyr/seo- — all sixteen show up. Pick seo-content-refresher, say "run this on pixelden.io", and Claude runs a structured refresh pass using ga4_compare_periods, gsc_search_analytics, and psi_analyze — the exact tools the playbook asks for, all live against your actual accounts.

No npx, no workspace config, no connector mapping.

2. Skill Contracts in every tool description

Shipping playbooks alone is not enough if the model does not know how tools chain. In 2.0, every SEO-relevant Metrifyr tool description now carries a Skill Contract block:

[Skill Contract]
Connector: ~~search console (tier 2)
Reads: your GSC property
Quality bar: cite exact impressions/position per query — not "improve CTR"
Next actions:
  • losing pages → psi_analyze on top 3
  • positions 8–20 → content_refresher playbook
  • low CTR >2k impressions → title/meta rewrite

The model sees the contract as part of the tool description. It does not need to guess what the tool returns or which tool to call next — the next-action hints are right there. The effect is a noticeable reduction in the "call a tool, get data, dead-end" pattern, because the tool itself tells the model where to go.

Contracts are registered centrally. Tools without a contract (internal housekeeping, job management) fall through unchanged.

3. Two new tools

knowledge_graph_lookup

A thin wrapper over the Google Knowledge Graph Search API. Ask "what does Google know about OpenAI?" or "is pixelden.io in the Knowledge Graph yet?" and get a structured answer: @id, @type, description, URL, image, score.

Why it matters for SEO and GEO:

Knowledge Panel eligibility — KG presence is a prerequisite for the panel. If Google does not have an entity for your brand, the panel cannot render.
AI Overview citations — Gemini and AI Overviews prefer entities they can resolve. A brand with a KG entity is a cleaner citation target than one without.
JSON-LD sameAs — once you find your entity's @id (e.g. kg:/m/0dl567), you can reference it from your site's JSON-LD via sameAs, reinforcing the link.
Disambiguation — if your brand name is generic, the lookup shows all candidate entities with their types and scores so you can pick the right one to optimise around.

knowledge_graph_lookup requires a GOOGLE_KG_API_KEY environment variable. Grab one from the Google Cloud Console — free tier is 100k queries per day. The tool only registers when the key is present, so if you are not using it, it stays off your tool surface.

schema_validate

Pre-flight validator for Schema.org JSON-LD. Point it at a URL or hand it raw HTML, and it:

Extracts every <script type="application/ld+json"> block
Parses each as JSON (reports parse errors with line context)
Checks required + recommended fields per @type — Article, NewsArticle, BlogPosting, Product, Organization, LocalBusiness, Person, WebSite, WebPage, BreadcrumbList, FAQPage, QAPage, HowTo, Recipe, Event, VideoObject, Review, JobPosting, SoftwareApplication, Book, Course
Returns a structured report: per-block @type, parsed data, and issue list categorised into errors (required-field violations, invalid JSON) and warnings (missing recommended fields, non-schema.org @context)

schema_validate is purely local. No external API, no fetch to Google, no quota. It runs fast enough to call on every page of a sitemap in a single skill invocation. It is the fast pre-flight check — for authoritative rich-results verification, Google's Rich Results Test UI is still the ground truth. This tool catches 80% of issues before you ever hit Google's tool.

4. Redesigned homepage and SEO surface

The homepage redesign shipped last week (d13c626) is rolled into 2.0 as the official release: tightened H1, expanded keyword coverage in meta and body copy, refreshed visual identity. If you have not looked at metrifyr.cloud for a few weeks, this is a fresh first impression.

What is next

The indexation stack is next. Two tasks in the backlog:

IndexNow integration — push notifications to Bing, Yandex, Seznam.cz, Naver, Yep via the IndexNow protocol. No OAuth, one API call, 10k URLs per request.
Google Indexing API — the missing push endpoint for Google. Not in IndexNow (Google does not participate), but the Indexing API works for general URLs with a caveat — it is officially for JobPosting and BroadcastEvent only, unofficially accepted for any URL at 200 URLs per day.

Together they give Metrifyr a complete push-indexation surface: new content → one tool call → Google, Bing, Yandex and Seznam notified within minutes.

Version bump

mcp-server 1.8.1 → 2.0.0
web → 2.0.0
MCP session lifetime is still 180 days (landed in 1.10.1)
Total tool count: 70+ live tools, 16 prompts, 5 resources

If you are already connected, nothing breaks — 2.0 is additive. New tools show up on the next session. If you want knowledge_graph_lookup, set GOOGLE_KG_API_KEY on your deployment and it registers automatically.

Feedback, bugs, requests — the per-tool feedback widget in the dashboard is the fastest path. I read every one.