Since v1.5, ZensInk shipped a zero-cost domain rating tool on Ahrefs' public API, brand-keyword fingerprinting so KD tools stop lying about branded queries, an MCP server so AI agents can drive the toolkit, and docs in four languages.
ZensInk shipped a lot since the v1.5 GEO Score release. The open source toolkit grew past 25 tools, the repo crossed 40 stars, and the whole pipeline is now something an AI agent can drive end to end.
This is a recap of what changed and the reasoning behind it.
Free domain ratings, zero API spend
The new domain_rating tool pulls real Domain Rating numbers from Ahrefs’ free public API. Batch input, CSV or JSON export, local caching, zero dependencies.
The catch with DR has always been the price: the number itself is free to look up, but the toolkit around it starts at $550 a month if you go through the usual route. We wanted DR as context, not as a product. So the tool costs nothing to run, and the only ask is attribution (“Domain Rating by Ahrefs”), which their license requires anyway.
A word of warning we keep repeating: DR is a weak signal on its own. We’ve audited sites with DR 37 built almost entirely on spam links, and sites with DR 8 that outrank them. DR tells you how big a domain looks. It says nothing about whether the links are real. Use it as a sanity check, never as a verdict.
Better keyword difficulty: brand detection and entry difficulty
Keyword difficulty scores from big tools have a structural problem: they blend seats you can never take. If the top 3 results are the brand’s own homepage, an official support doc, and a YouTube video, the “difficulty” number is meaningless for you, because those seats aren’t contestable.
The upgraded kd module now handles this in three passes:
- Brand-keyword fingerprinting. Three fingerprints: official domain in top 3, two or more domains from the same brand family on page one, and platform ecosystem density (Reddit, YouTube, docs sites stacking the results). Two or more hits means the query is effectively a brand query, and generic KD should be discarded.
- Entry difficulty (kd_entry). Recalculate difficulty after removing official pages and fixed platform seats. What’s left is the difficulty of the seats you can actually fight for.
- Link budget estimation. Translate difficulty into a rough number of referring domains an editorial page needs, via a curve interpolation. Not a promise, but a budgeting aid: if the entry difficulty implies 80 referring domains and you have 6, the math is telling you something.
This came out of studying how SERPs actually behave in 2026, where AI Overviews and platform pages absorb more of page one every quarter. If your KD tool doesn’t separate contestable seats from fixed ones, it is lying to you, politely, every day.
Agent-ready, measured
Two things shipped on this front:
- MCP server built in.
python3 -m zens_ink.mcpexposes every CLI tool over stdio. Point Claude, Cursor, or any MCP client at it and the agent can run keyword research, audits, and rank tracking without you typing commands. - SKILL.md everywhere. Every package, free and Pro, ships with an agent instruction file, so tools describe themselves to whoever (or whatever) is driving.
We also ran zens.ink itself through an agentic-readiness audit: the site scores 98/100 on is-agentic.com, up from 64 when we first measured. That case study is public, warts and all, at zens.ink.
The split is deliberate, same as always: the free layer tells you whether you have a problem. The Pro workflow engine tells you what to do about it, ranked by what your own GSC data says you can win.
Docs in four languages
The documentation site is now fully translated into Japanese and Spanish, joining English and Chinese. Nine long-form docs per language, no machine-placeholder pages. Japan and Spain are two of the largest developer markets where “free-tier SEO tooling” is a genuinely under-served niche, and translated docs consistently outperform translated marketing pages, so the docs came first.
The CLI also installs straight from git now:
pip install git+https://github.com/ZensInk/zens-ink-seo-package.gitHousekeeping
The repo moved to the ZensInk org on GitHub. Old links redirect, stars and forks came along. The project sits at 44 stars and 13 forks as of this writing, which for a niche Python SEO toolkit feels about right: small, but the right people.
What’s next
- GEO Score calibration continues against real sites, layer weights still adjusting
- Winability refinements, feeding more GSC history into the personalized difficulty model
- A Mac-native workbench is in early scoping: the audit engine wrapped in a desktop app, for people who want the report without the terminal
Build log, not a press release. Numbers are real as of today. If something breaks, tell us. We fix things.