标签：#seo

I do a fair amount of competitor backlink research, and the workflow always annoyed me: open a dashboard, run a query, export a CSV, eyeball it, copy domains into a doc, switch to email. Lots of tab-hopping for what is fundamentally a data-filtering problem an agent should handle. So I wrapped the backlink API I'd been using into an MCP server. Now I stay in Claude Code (or Cursor, Cline, Zed, Windsurf) and just describe the goal. This is the build: the architecture, the four tools, and the one design decision I'm still not sure about. The data source The server runs on the Common Crawl hyperlink webgraph — about 4.4 billion edges across 120 million domains, published quarterly as Parquet. That matters for an MCP tool specifically: the data is open, so there's no scraped-proprietary-index liability in handing it to an agent, and the same query is reproducible by anyone. The HTTP API in front of it ( CrawlGraph ) does the heavy DuckDB work; the MCP server is a thin TypeScript stdio client over it. Keeping the server thin was deliberate — all the query cost, caching, and quota logic lives server-side, so the MCP package stays a ~300-line wrapper that's easy to audit before you hand it your API key. The four tools backlinks → referring domains for a target, with authority scores gap_analysis → domains linking to your competitors but not to you gap_outreach_targets → the composite play (below) releases → list the Common Crawl snapshots backlinks and gap_analysis map 1:1 to API endpoints. gap_analysis is the interesting primitive: submit your domain plus 2-5 competitors, and it returns every domain that links to at least one competitor but not to you, each tagged with a found_on array listing which competitors it links to. The composite tool, and the decision I'm unsure about Most API-wrapper MCP servers are pure 1:1 mappings. I added one opinionated composite tool, gap_outreach_targets , because the raw gap output isn't the thing you actually want — it's the raw materia

In web development, every millisecond on the critical path matters. Typography is part of your visual identity, yet a single self-hosted family can add hundreds of kilobytes before the browser paints the hero line clients actually see. That cost often shows up as slower First Contentful Paint (FCP) and a worse Largest Contentful Paint (LCP) when the LCP element is a headline set in a custom face. Font subsetting is the practical fix: ship only the characters (glyphs) each page or component needs instead of entire typefaces built for every language and symbol you will never render. What font subsetting is A typical desktop font file contains thousands of glyphs: Latin extended, Cyrillic, ligatures, numerals, punctuation, and symbols you may never use on a marketing site. When you @font-face that file or pull it from a CDN, the browser still has to download and parse the whole package unless you split it. Subsetting creates a smaller font file that includes a defined subset of Unicode code points. Examples: A display font used only on one H1 might need fewer than 40 characters. A Latin-only blog body font does not need CJK tables on every article. UI labels in English can omit unused weights and scripts. The browser downloads less data, spends less time decoding, and can apply the face sooner. You keep the look; you drop the dead weight. When font subsetting improves web font performance and LCP Subsetting pays off fastest in these cases: Situation Why subsetting helps Display or hero typography Few unique characters, high visual impact, often on the LCP text node Single-language sites Remove unused scripts and diacritic ranges you do not publish Icon or logo fonts misused as full fonts Replace with SVG where possible; subset if you must keep a font icon set Multiple weights loaded globally Subset per weight, or load weights only on routes that need them It is lower priority when you already use a system font stack for body copy and only load one small variable font w

An account manager forwarded us a Search Console screenshot at 4:47 p.m. on a Friday. LCP on the homepage had been in the red for eleven days. Engineering had not been in the thread because nobody was assigned to watch the numbers until a sponsor noticed. That is reactive performance monitoring in practice: fast tools, slow humans, and a calendar that only opens when someone else raises the alarm. Proactive monitoring is the opposite shape. Scheduled lab tests run whether or not anyone remembered to open PageSpeed Insights. Performance budgets turn history into rules. Alerts fire when those rules break, with enough context to triage and enough restraint that people still read them. The word “smart” in product copy usually means the second part: alerts that match how teams actually work, not a louder siren. This post is for agency leads who already automate some tests but still learn about regressions from client inboxes. We walk through what reactive looks like, what proactive requires, and how to wire alerts so they change behaviour instead of training everyone to mute the integration. What reactive performance monitoring looks like on agency teams Reactive does not mean “no tools.” Most agencies we speak with have PageSpeed Insights bookmarks, a monthly spreadsheet, and a standing item on the retainer review slide. Reactive means the signal arrives after the business event you wanted to prevent. Common patterns: Spot checks after deploy. Someone runs PSI on staging, ships, and assumes production matches until a complaint arrives. Quarterly or monthly report pulls. CrUX and Lighthouse exports go into a deck. Useful for storytelling; too slow to catch a plugin update that moved INP on Tuesday. Client-led discovery. The sponsor pastes a screenshot, asks “is this new?”, and the team scrambles to reproduce on a laptop that may not match the lab conditions you use in production monitoring. Alerting without policy. Webhooks fire on every metric twitch. By week three the

Originally published at https://rnye.tech Hi all, today's post details an interesting problem that faced a website thanks to undocumented Google crawl behaviour that hit us suddenly. The website used Azure Front Door for global CDN/WAF capability but only had one origin - hosted in Azure in the UK South region. This should have been fine given it's a UK-centric site that receives very little global traffic - that is until Google starts crawling you from the West Coast of the US suddenly. Let's dive in. The problem: what Google Search Console was telling us All was well with the site from the UK, cache hit ratios were in the 80%+ range, response times were generally rapid even if cache was missed. Google typically crawled the site thousands of times a day. Then suddenly a ticket came in detailing a drastic drop-off in mid-April. Average response times were never great according to Google (700ms) but they'd suddenly jumped to almost double that number (1.3s) with seemingly no explanation. There was absolutely no denying the correlation between crawl requests and average response time, and indeed this is documented behaviour - if response times increase, Google backs off. They claim it's to prevent overloading the site, and I believe that, but I also feel it's likely to ensure they're not wasting their crawl compute resources on long-loading pages. Either way, Google Search Console offered zero explanation as to why. So the team did what most dev/devops teams do - review latest changes, any Azure Front Door configuration changes in particular, as well as wider site changes. And nothing correlated. AFD hadn't been changed for two weeks, it was that stable, and other changes weren't remotely related. Besides, genuine traffic in the UK wasn't seeing the impact. Response times were still good in the P50/P90/P99 metrics. Tracing the cause with Azure Front Door logs and AI As I've touched on in my last post , AI can be fantastic at quick data analysis. That's not to say a hu