@ouranos-labs
基本信息
SpamBrain-proof your pSEO before you publish, audit page relationships, not just pages. AI-triaged findings, 8 LLM providers, local telemetry, cost caps, MCP server.
使用下面的配置,将此服务器添加到你的 MCP 客户端。
{
"mcpServers": {
"pseolint": {
"command": "npx",
"args": [
"@pseolint/mcp"
]
}
}
}Use when a user asks to check their website for SEO issues, SpamBrain risk, duplicate content, thin pages, or before deploying a programmatic SEO site. Crawls the site, runs 40+ rules scored across 4 categories (integrity, discoverability, citation, data), and returns a verdict (ready/caution/concerning/critical) plus a numeric risk score (0-100, lower is better) with actionable findings. Pre-flight site classification suppresses pSEO-targeted rules on small marketing sites and blogs unless strict mode is set. When the audited site has ≥2 detected URL templates, the response includes a `templates` array with per-template verdicts, risk scores, category grades, variance metrics (uniformityScore, topDriver), and audited-URL counts. Both human-readable text and structured data (structuredContent) are returned. A high-risk site is NOT a tool error — check the `passed` field / `verdict`, not `isError`.
Use when a user wants to understand WHY their pseolint verdict is concerning/critical, what categories are failing, and what to fix first. Returns a prioritized breakdown with quick wins listed before structural fixes, plus a pass/fail verdict against the risk threshold. Both human-readable text and structured data (structuredContent) are returned. A failing threshold is reported via `passed`, not `isError`.
Use when a user asks to check a specific page URL for technical SEO issues. Checks per-page rules only: canonical tags, Open Graph tags, JSON-LD schema, robots directives, meta tags, thin content, and author signals. Does NOT check cross-page rules (duplicates, cannibalization, linking) — use pseolint_audit_site for those. Both human-readable text and structured data (structuredContent) are returned.
Use when a user wants concrete, paste-able fixes (rewritten H1s, JSON-LD blocks, robots.txt patches, internal-link suggestions) — not just a list of issues. An LLM drives 25 tools (sitemap fetch, template clustering, per-page rule checks, AEO probes) and produces a fix manifest with structured patches, each validated against a deterministic schema. Costs real money (~$1-3 per audit on managed Anthropic). Capped at $2 / 60 tool calls / 180 seconds by default — adjust if the user asks for a deeper run. Returns a text summary plus a compact structured summary (structuredContent); pass format:'json' to also get the full manifest in the text content. The structured `reason` field reports why the run stopped — an incomplete run is not flagged via isError. The full manifest can be streamed via the CLI: `pseolint orchestrate <domain> --manifest-out manifest.json`.
pseolint is an MCP server that audits programmatic SEO sites by template rather than by URL. It detects graph-level issues such as doorway clusters, near-duplicates, and entity-swap patterns across thousands of pages, providing per-template verdicts and variance metrics. Designed for developers and SEO professionals, it integrates with Claude and coding agents via its skills suite.
Run npx pseolint http://localhost:3000 to audit a site. Add skills for Claude with npx skills add ouranos-labs/pseolint --skill pseolint aeo. The MCP server provides rule knowledge as resources and supports parameters like authorityScore and --render for advanced checks.
--ci-threshold.pseolint works at the graph level, not per page. It detects doorway patterns, near‑duplicate template clusters, and entity‑swap doorways that cross thousands of URLs—issues invisible to editorially‑focused tools.
pseolint is a Node.js package. The CLI requires Node 18+; the MCP server and render‑aware checks (`--render`) need Playwright installed. It is distributed via npm.
No. pseolint is authority‑blind by design. It does not fetch Moz, Ahrefs, or Semrush data. You can pass a manual `authorityScore` (0‑100) via CLI flag or MCP param to adjust verdict leniency.
It does not detect Core Web Vitals, image SEO, Open Graph completeness, title‑tag uniqueness, H1 structure, schema‑content drift, outbound link health, search intent, or parameter crawl‑budget waste. The full list is documented at pseolint.dev/blind‑spots.
It uses a two‑phase pipeline: first clusters the sitemap into templates (cheap fetches), then deeply audits K sampled pages per template. Typical budget is ~80 fetches on a 100k‑URL site, covering every template.
Request dedicated senior developers to work on your project, write job descriptions, and prep technical interviews — all without leaving your Claude chat. Just describe what you need, and Lemon.io MCP does the rest. 1–3
PickFu MCP is a Model Context Protocol (MCP) server that connects AI assistants, developers, marketers, product managers, and AI agents to PickFu's consumer research platform. It enables users to create surveys, conduct
A good sequence diagram can solve almost any communication problem in engineering. Volley lets your agent write them as plain text, with the real request and response payloads on every arrow. It checks its own syntax wi
Handl speaks the Model Context Protocol. Your whole billing operation — projects, invoices, clients, reports, approvals — as tools any AI assistant can call, inside your autonomy rules.
See and control a real Windows PC you own — from any MCP client, locally or remotely. Observe (UIA + screenshots), click/type/drag/scroll, launch apps, owner Live View. BYOH: your machine, your key.
评论