Skip to content

Pagerender MCP server

@pagerender/mcp gives an MCP client 22 tools that read and 14 that edit page content. 5 of them need no token and no account.

Sign in with your browser

Add the hosted URL to your client and sign in to Pagerender. You choose read only or read and edit on the sign-in page. The access covers every domain on the account.

Connector URL
https://app.pagerender.io/mcp

Claude and Claude Desktop: Settings > Connectors > Add custom connector, paste the URL, sign in. Other clients:

Terminal
claude mcp add --transport http pagerender https://app.pagerender.io/mcp

Run /mcp in Claude Code and choose Authenticate.

Use an API key instead

For setups without a browser. The server runs on your machine as npx -y @pagerender/mcp. Remove the token line to use only the tools that need none.

Terminal
claude mcp add pagerender \
  --env PAGERENDER_TOKEN=your_token \
  -- npx -y @pagerender/mcp

Environment variables

The key-based server reads these once, at startup. Signing in through the hosted URL needs none of them.

  • PAGERENDER_TOKENAn API key, created under Security in the dashboard. Use a read-only key to read, or a content key to also edit titles, descriptions and meta rules. Your client stores it in plain text, so never use a full-access key here.
  • PAGERENDER_MCP_WRITESSet to true to add the tools that write content. Leave it unset and the server only reads.
  • PAGERENDER_API_URLOverrides the API base URL. Defaults to https://api.pagerender.io.

22 tools that read

  • analyze_urlWhat a crawler extracts from a public URL: title, canonical, headings, word count, crawlable links versus JavaScript-only links, and whether the response is an empty app shell.No token
  • analyze_siteSamples up to 7 pages from the URL’s own links and returns a verdict for each, plus which AI crawlers robots.txt blocks. Runs on your machine.No token
  • check_ai_crawlersWhich of 21 AI crawlers a domain blocks sitewide in robots.txt, and which it allows.No token
  • service_healthQueue depth, cache and renderer health for the Pagerender service.No token
  • list_plansPlans and limits published by Pagerender.No token
  • verify_integrationConfirm a URL is served rendered HTML rather than the raw JavaScript shell.Needs a token
  • render_statsRender volume and cache hit rate, filtered by period and domain.Needs a token
  • index_statusWhat happened to one URL in the indexing agent, its last Google verdict, and what the agent does next.Needs a token
  • indexing_overviewThe indexing agent for one domain: recent runs, submit and skip decisions, index rate, remaining quota and stalled patterns.Needs a token
  • list_pagesPages of a domain with their state, filtered and sorted.Needs a token
  • list_page_sectionsThe sections of a domain, with how many pages each holds.Needs a token
  • lookup_pageOne page: its served title and description, where each came from, and its index state.Needs a token
  • list_meta_rulesThe URL pattern rules that set titles, descriptions and canonicals.Needs a token
  • preview_meta_ruleHow a rule would change the pages it matches, before you add it.Needs a token
  • list_meta_changesRecent title, description, link and schema changes, newest first.Needs a token
  • list_link_recommendationsInternal links Pagerender suggests adding.Needs a token
  • preview_llms_txtThe llms.txt Pagerender serves for a domain.Needs a token
  • list_actionsThe settings actions Pagerender can run, with their arguments.Needs a token
  • preview_actionsWhat settings actions would change, before and after. Writes nothing.Needs a token
  • settings_changesRecent settings changes with their ids.Needs a token
  • match_page_factsMatches a sheet of page facts to pages by URL, by an ID in the URL, or by name.Needs a token
  • meta_run_statusThe meta writer’s latest run and every page it wrote.Needs a token

14 tools that edit content

They need a content key and PAGERENDER_MCP_WRITES=true, or a connector you allowed to edit content. They change what your site serves to crawlers, or spend AI allowance.

  • start_meta_runWrites titles and descriptions for a section, optionally from a sheet of facts.Content key
  • apply_meta_runSaves a run’s titles and descriptions so they serve on the next render.Content key
  • set_page_metaSets one page’s title and description.Content key
  • suggest_page_metaThe AI’s title and description for one page, without saving.Content key
  • rewrite_page_metaRewrites a title or description: shorter, with a keyword, with the brand, or as asked.Content key
  • add_meta_ruleAdds a rule for a URL pattern.Content key
  • remove_meta_ruleRemoves a rule.Content key
  • suggest_meta_ruleThe AI’s rule for pages missing a title or description.Content key
  • set_brandSets or clears the brand added to titles.Content key
  • undo_changeUndoes a title, description or schema change.Content key
  • regenerate_link_recommendationsRecomputes internal link suggestions.Content key
  • apply_actionsRuns undoable settings actions. Final actions such as a cache purge are refused.Content key
  • undo_settings_changeUndoes a settings change.Content key
  • redo_settings_changeRedoes an undone settings change.Content key

render, purge and warmup are left out. They spend render quota or change what is cached, and an MCP client has no confirmation step for them. Use the CLI or the API for those.

What it costs

Nothing, and there is no quota. The free tools fetch the site from your own machine, so they never touch our servers.

  • analyze_url, one pageNo limit.
  • analyze_site, up to 7 pagesNo limit.
  • check_ai_crawlers, one domainNo limit.
  • pagerender audit, any number of pagesNo limit.

What an answer looks like

transcript
You: is example.com's content visible to Googlebot?

Agent calls analyze_url, then answers:

  No. The page returns an empty app shell: 12 words in the body,
  a div with id="root" and no crawlable links. Googlebot renders
  JavaScript on a second pass, so it may index it eventually.
  The AI crawlers behind ChatGPT and Perplexity will not.

We use optional analytics cookies to see how visitors find Pagerender and which parts of the product they use. Privacy policy