Make your coding agent understand how your website should be discovered — then let it fix the code.
RankCore gives AI coding agents (like Claude Code, Cursor, Codex, Gemini CLI, and Cline) a current search/AI-search workflow paired with a deterministic local audit engine.
Instead of relying on heuristic SEO advice from outdated data, RankCore executes a rigorous, real-time technical audit of your project and presents evidence-based fixes directly to your agent.
Status: Early development. The
/rankskill and the deterministic audit engine work end to end; see the master plan for the roadmap.
- Zero-Key: No RankCore account or external SEO APIs required.
- Local Native Binary: Built in Go. No Node or Python runtime requirements.
- Deterministic: Only enforces strict, schema-first rules. No arbitrary SEO scores.
- SSRF Protected: Safe HTTP crawler with loopback protections built-in.
- Bounded concurrent crawler: Same-origin frontier with page, depth and body-size limits, include/exclude patterns,
robots.txtetiquette for remote sites,429/503backoff, and SSRF protection (validated-IP dialing, every redirect hop re-checked). - Sitemap-aware discovery: Reads
Sitemap:lines from robots.txt (or/sitemap.xml), follows sitemap indexes and gzip, and crawls every listed URL alongside linked pages. - 35 deterministic rules across availability, indexability, canonicals, links, metadata, content, internationalization, structured data, sitemaps, renderability and AI-crawler access. Highlights:
- 🛑
4xx/5xxpages, redirect chains and loops,noindexin meta orX-Robots-Tag - 🛑 Broken and redirecting internal links, uncrawlable anchors, orphan pages, deep pages
- 🛑 Canonical targets that 404, redirect, are noindexed or chain; conflicting canonicals
- 🛑 Sitemap entries that error, redirect, are noindexed or non-canonical
- 🛑 Duplicate titles and descriptions, hreflang without return links
- 🛑 JSON-LD syntax errors and missing required properties for Google rich-result types
- 🛑 Search and AI-search crawlers blocked in robots.txt (training crawlers are reported but never flagged)
- 🛑
- Verification diffing: Re-audits and classifies every baseline finding as fixed, still present, regressed, new or not re-testable, with CI-friendly exit codes.
- Rendered mode: Compares server HTML with the DOM rendered by an already-installed Chrome, Edge, Brave or Chromium.
- Local builds: Point
rankcore auditat a static output directory and it serves it on a loopback port.
Run rankcore explain for the full rule catalog.
RankCore installs directly from this GitHub repository — no custom domain, account, or API key required. The installer downloads the latest precompiled release from GitHub Releases, verifies its SHA-256 checksum, installs it to a user-writable directory (no sudo/admin), and runs rankcore setup to register the /rank skill with your detected coding agents.
curl -fsSL https://raw.githubusercontent.com/zimkk/rankcore/main/install.sh | shirm https://raw.githubusercontent.com/zimkk/rankcore/main/install.ps1 | iexRequires Go 1.26+:
go install github.com/zimkk/rankcore/cmd/rankcore@latestNote: Precompiled binaries are published on tagged releases. If the installer reports a missing release for your platform, use the source build above or download manually from the releases page.
After installation, RankCore automatically registers itself with your detected agents (e.g., ~/.cursor/skills/rank, ~/.gemini/skills/rank).
Open your web project in your coding agent and type:
/rank
The agent will seamlessly:
- Understand your project topology.
- Run a deterministic local technical SEO audit.
- Recommend fixes based strictly on evidence.
- Apply the fixes through your agent and verify them.
You can also use RankCore manually from your terminal:
# Check environment health
rankcore doctor
# Run an audit on a local or remote target, or a static build directory
rankcore audit https://your-site.com
rankcore audit http://localhost:3000 --render auto --max-pages 200
rankcore audit ./dist
# Machine-readable output (stdout is pure JSON) and a CI gate
rankcore audit https://your-site.com --json > audit.json
rankcore audit https://your-site.com --fail-on high
# Verify fixes against a baseline
rankcore verify https://your-site.com --baseline .rankcore/runs/latest/audit.json
# Explain a rule
rankcore explain RC-CANON-002Exit codes: 0 success, 1 operational error, 2 gate failure (--fail-on threshold hit, or verify found regressions, new findings, or unresolved critical issues).
RankCore is built on a clean, scalable Go module architecture:
internal/crawl: The concurrent frontier, URL normalization and page snapshots.internal/httpx: SSRF-safe HTTP client with manual redirect tracking and backoff.internal/extract: HTML parsing for metadata, links, headings, images, JSON-LD and visible text.internal/robots,internal/sitemap,internal/aiaccess: RFC 9309 robots, sitemaps, and the AI crawler registry.internal/rules: Page rules and site-wide rules (link graph, canonicals, sitemaps, duplicates).internal/audit,internal/verify: The audit pipeline and baseline diffing.internal/report: Golden JSON schemas forAuditReportandFinding.internal/setup: Agent integration and skill installation.
We welcome contributions! Please review our Contributing Guidelines and Security Policy before submitting pull requests.
To run the test suite locally:
go test ./...This project is licensed under the MIT License - see the LICENSE file for details.