aicodereview.io

An independent directory of AI code review tools, each scored against a public 9-pillar standard, plus head-to-head comparisons, guides, a glossary and a blog. Funded by Kodus, who build one of the listed tools and are scored by the same rubric.

Built with Astro 5 + Tailwind 4, deployed on Vercel. Pages are prerendered (static HTML on the CDN); only /api/* runs on demand as a serverless function.

Structure

src/
├── content/
│   ├── tools/     # one JSON file per tool — the dataset behind every score
│   ├── pillars/   # the 9 standards (MDX)
│   ├── glossary/  # glossary terms (MDX)
│   ├── learn/     # long-form guides (MDX)
│   └── blog/      # blog posts (MDX) — by hand or via the content API
├── lib/
│   ├── score.ts   # the scoring rubric: statuses → weights → coverage score
│   ├── tools.ts   # directory ordering, labels, derived pros/cons
│   ├── compare.ts # which pairs get a page, and the generated verdict prose
│   ├── blog.ts    # blog helpers, tags, related posts
│   └── glossary.ts
└── pages/
    ├── index.astro               # the directory (home)
    ├── tools/[slug].astro        # tool profile (+ .md mirror)
    ├── tools/category/…          # category, self-hosted and open-source views
    ├── compare/[pair].astro      # generated head-to-head comparisons
    ├── standards/                # the 9 pillars + per-pillar market coverage
    ├── glossary/, learn/, blog/  # reference and editorial
    ├── methodology.astro         # the rubric and the funding disclosure
    ├── api/posts.ts              # content API (see CONTENT-API.md)
    └── llms.txt.ts, llms-full.txt.ts, rss.xml.ts, search-index.json.ts

Commands

Command	Action
npm install	Install dependencies
npm run dev	Dev server at localhost:4321
npm run build	Build (static pages + .vercel/output with the /api function)

Scoring

Each tool is mapped against the 9 standards with four statuses: documented (1 point), partial (0.5), not offered (0), undocumented (0). The sum is the coverage score out of 9, and it drives the directory order, the comparison tables and the "best on this standard" lists.

Undocumented scores zero deliberately — a capability a buyer cannot verify is one they cannot rely on — and every profile shows how many pillars are undocumented so a low score can be read correctly. The full rationale is on /methodology/.

Adding or correcting a tool means editing one JSON file in src/content/tools/. Every field must be readable from a primary source (the vendor's pricing page, docs or repository) and stamped with lastVerified; fields with no public source are left unset rather than estimated.

Content API

Programmatic publishing (POST /api/posts, git-backed): see CONTENT-API.md. Copy .env.example to .env and set CONTENT_API_KEY to use it locally.

SEO/GEO notes

• Structured data per page type: ItemList on the directory and every filtered view, SoftwareApplication + Offer on tool profiles, DefinedTerm/DefinedTermSet on the glossary, Article on posts and guides, FAQPage wherever there is an FAQ, plus BreadcrumbList and a site-level Organization/WebSite graph.
• /llms.txt and /llms-full.txt expose the whole site to AI engines; robots.txt explicitly welcomes AI crawlers. Raw markdown mirrors exist for posts (/blog/<slug>.md), tool profiles (/tools/<slug>.md) and glossary terms (/glossary/<term>.md).
• Comparison pages are generated from the dataset for a curated set of pairs (src/lib/compare.ts); pairs that already have a hand-written article are skipped so the two never compete for the same keyword.
• /tools redirects to /, which is the directory — one canonical page for the primary keyword.