Reproleaf

SEO content checker and internal linking tool

Review an article, inspect its headings and explore internal links in one workspace. Paste a draft or import a WordPress inventory for clear checks and practical next steps.

Reproleaf · Free workspace · Article / Site inventory / MCP

SEO workspace

Paste a Markdown draft to get strengths, issues and prioritized edits. The checker runs in your browser without sending the draft to the tool server.

A practical SEO content checker before publication

Find your next edit instead of chasing an unexplained score. Reproleaf parses a Markdown draft, compares headings and links with supplied records, and separates checked items, review points, structural problems and missing data. Use it before moving an article into your CMS or as a second review of an existing draft.

Page titles, search queries and heading structure

Add the search title, meta description, primary query and actual template H1. The panel shows H1–H6 structure, skipped heading levels and exact phrase occurrences. When WordPress supplies the H1, a second H1 in the body is unnecessary. An identical title and H1 is not automatically an error. Word counts and phrase counts are observations, not instructions to meet a keyword-density target.

Evidence, terminology and image descriptions

For a deeper editorial workflow, supply a JSON packet with declared claims, source records, glossary rules and calculation reproduction steps. The checker identifies missing bindings and conflicting terminology. Images with missing alt text are review candidates: decorative images can legitimately use an empty description. Record checks do not verify source truth or replace an editor reading the cited material.

Use your stack, not a required CMS

Markdown and the common JSON format work with Astro, Nuxt/Vue, React/Next, static generators and custom sites. A local exporter reads an explicitly selected HTML build. The WordPress plugin is an additional adapter, not a platform requirement.

node integrations/static-site/export.mjs --dir ./dist --site https://example.com --out inventory.json

For Astro, point to the static HTML output; use the published build directory for other generators. The exporter does not execute JavaScript. A client-rendered SPA without article HTML needs a content-system or application-data export.

Import published pages from the free WordPress export helper. Explore observed relationships, distinct linking pages, destination counts and articles with no incoming link in the supplied set. Review potential internal linking opportunities based on overlapping title and supplied-text terms. Suggestions are candidates for editorial review, not automatic link insertions.

Understand what a sitemap can tell you

An XML sitemap lists page URLs, not the hyperlinks connecting them. Import it as a URL inventory. To build a link graph and recommend related pages, supply a stored-content JSON export instead. If you have a sitemap index, upload one of its child URL sitemaps with a urlset root. The current workspace accepts up to 1000 pages and files up to 5 MiB.

Topic clusters and site architecture

Supplied category labels help review links within a topic cluster. A missing topic connection is an invitation to inspect reader navigation, not a rule to force a link. The export can omit menus, widgets and dynamic blocks. No incoming link in this set is an orphan-page candidate, not proof of a site-wide orphan. Outgoing external links are not backlinks to your domain.

Who is this workspace for?

  • Writers and editors: inspect a draft and build a focused editing checklist.
  • Blog owners: find weak connections between published resources.
  • SEO practitioners: add an inspectable structural report to search-intent and SERP research.
  • Developers and AI agents: call the same checks through local MCP, CLI or HTTP API.

Why use Reproleaf?

The browser workspace and machine-readable reports share the same checking core. Inspect the findings, download the JSON and repeat the analysis. Discuss a selected report with ChatGPT, Perplexity or Gemini when you want another perspective. Browser checks keep the draft on your device. There is no account requirement or paid model API key. The new tool implementation is available in a downloadable MIT-licensed source archive; private project source is excluded.

SEO MCP tools, a local API and an embeddable checker

Download the kit and install dependencies with Node.js 22 or later. It includes eight read-only MCP tools and three skills: mobile-web-review, seo-evidence-review and blog-link-review. The current MCP server uses local stdio. Opening an AI chat does not install it.

Download source and MCP ↗WordPress export helper ↗Self-host the workspace ↗
npm ci --ignore-scripts
node src/mcp.mjs
# Or run the local HTTP API:
node src/api.mjs

The local API accepts JSON at POST /v1/editorial and POST /v1/blog. Host the static workspace on your own HTTPS domain, then embed it with an iframe. Setup instructions and limits are in the README. A public hosted API is not available yet.

Mobile web regression checks for agents

The kit also includes separate minimal Vue, React, Three.js, Phaser and HTML reproductions for modal close controls, scroll interactions, canvas coverage and overflow. These verify observable behavior after a change. Chromium viewport emulation does not certify a physical iPhone.

Connect your agent

Choose your client. Editorial checks and frontend cases share one toolkit; each client has its own connection settings.

Codex

Extract the kit, install Node.js 22+ and run npm ci –ignore-scripts in the kit directory. Replace the example with your absolute path. Commands register MCP in your client; this page does not execute them.

codex mcp add reproleaf -- node "C:/tools/reproleaf/src/mcp.mjs"
codex mcp list
Official documentation ↗

After connecting, ask the agent to call list_workflows or list_cases. Use review_editorial_packet and audit_blog_links for editorial work; get_case and get_case_source for frontend cases. The toolkit does not automatically gain access to your projects or upload their source.

Data handling and limits

Drafts and inventories are processed in the browser. Keep credentials, customer personal data and confidential documents out of shared inputs. External AI handoff is separate: preview the prompt, then paste and submit it in the chosen service. A green status describes a specific check; it does not predict rankings, traffic or factual accuracy.

Need rendering, indexation and performance work across a site? Explore technical SEO services, the developer background and the blog.

Questions about the checker

Does this replace Yoast SEO?

Reproleaf checks supplied drafts and inventories. It does not manage canonical URLs, robots rules, sitemaps or WordPress metadata. Yoast can continue to own those settings.

Can a sitemap reveal internal linking opportunities?

XML supplies URLs. Observed incoming and outgoing links require a stored-content inventory. A sitemap contains neither article text nor proof of indexation.

What do the status colors mean?

Green means a specific check passed, amber needs review, red marks a structural problem, and gray means the input lacks data. These are not ranking predictions.

Is Reproleaf free?

The browser workspace, WordPress export helper and local toolkit are free. Third-party AI services have their own access requirements and terms.

Which sites and frameworks does Reproleaf support?

Content checks accept Markdown and a common JSON inventory independently of the CMS. A local exporter supports static HTML builds from Astro and other generators. WordPress is one adapter. UI checks include Vue, React, Three.js, Phaser and plain HTML cases.

Do I need a different MCP server for each framework?

No. One Reproleaf MCP server exposes the workflows, searchable cases and checking tools. The agent selects relevant material. Vue, React and editorial checks do not require separate MCP installations.

How do MCP, skills, the API and the workspace differ?

MCP exposes tools and reference material to an agent. A skill describes the working process and limitations. The API provides programmatic access to the same core. The workspace supports manual inputs and reports. These are different interfaces to one toolkit.

Is it a model trained on projects that prevents every bug?

No. It is a toolkit of generalized working rules and separately written reproducible examples, not a separately trained model or a guarantee. Cases help define an expectation and verify a fix. Relevance must be checked for each project, and failures outside the current coverage can remain.

Do I have to upload my project source?

No. Browser checks process the supplied draft and inventory locally. The local CLI and MCP use explicitly supplied inputs and bundled materials. Private client project source, keys and databases are not included in the public toolkit.

Can I connect Claude, Codex, Gemini, ChatGPT or Grok?

Local clients have stdio MCP setup instructions. Web clients require a reachable remote MCP endpoint or a supported tunnel. A public Reproleaf endpoint is not yet available, so explicit report handoff is available for those chats. Check the tab for your client.

Does the workspace already run a generative AI model?

Structural checks currently run through the local core, not a generative model. You can hand a report to your chosen AI chat. A hosted AI review with an agreed provider and daily spending limit is planned separately and is not enabled yet.

How can I use a report with my team or client?

Download JSON for an agent or an HTML report for reading and printing. HTML attributes the report tool to Reproleaf. Downloads are not automatically published. Review reports for private URLs and data before sharing.

Can I embed the workspace in my site?

Yes. Host the static workspace on your own HTTPS domain and embed it with an iframe. The local HTTP API is for your integrations and binds to 127.0.0.1 by default. Setup and examples are included in the archive.

Does the WordPress adapter edit articles or insert links?

No. It analyzes an inventory of published content in a bundled local admin workspace and exports JSON. It does not edit posts, replace Yoast or inject hidden backlinks.

Will the case library grow?

New cases can add separately written reproductions and independent acceptance checks. Search, filters and pagination keep the catalogue compact as it grows. Case count alone does not establish coverage quality.