Skip to content
Siteory

Siteory MCP

Siteory MCP and agent skills

Run Siteory audits from Claude Code, Cursor, Codex, VS Code, or any MCP client. Your agent starts the audit, reads the findings and their evidence, fixes your code, and re-scans to confirm. Siteory MCP, the agent skills, and the local CLI come with paid plans, from a $9 single audit. You sign in once in the browser. There is no API key to copy.

https://api.siteory.co/mcp
Connect your clientRun a free audit

What it does

Siteory MCP is a remote MCP server. It gives your agent ten tools that cover the whole loop, from the first audit to proof that a fix landed.

  1. 01

    Audit

    start_audit runs a full SEO, AI visibility, and security audit of a URL you own. quick_audit runs a quick preview that does not use a credit.

  2. 02

    Fix

    list_issues ranks the findings. get_issue returns the evidence, the affected URLs, and a fix prompt for your agent.

  3. 03

    Verify

    rescan runs the audit again after the change. compare_audits lists what was resolved, what is new, and what is still open.

Hosted audits fetch public URLs from Siteory's servers, so they cannot reach localhost or a private network. For those, use the local CLI. Only audit sites you own or are allowed to test.

Connect your client

Every client uses the same server URL and signs in through your browser with a Siteory account on a paid plan. Pick yours below.

Set up every agent at once

Finds Claude Code, Codex, Cursor, VS Code, Gemini CLI, OpenCode, and Windsurf on this computer, adds the Siteory server to each, and installs the agent skills. Add --dry-run to see the changes first. To print the config for one client instead, run npx siteory mcp.

Terminal

npx siteory setup

Claude Code

Add the server from a terminal. --scope user makes Siteory available in every project.

Terminal

claude mcp add --transport http --scope user siteory https://api.siteory.co/mcp

Then run /mcp in Claude Code, select siteory, and choose Authenticate.

Cursor

Add Siteory to ~/.cursor/mcp.json for every project, or to .cursor/mcp.json in one repository.

~/.cursor/mcp.json

{  "mcpServers": {    "siteory": {      "url": "https://api.siteory.co/mcp"    }  }}

Cursor lists siteory in its MCP settings. Sign in from there when it asks.

Codex

Add the server to your Codex config, then sign in. Codex uses $CODEX_HOME/config.toml instead when that variable is set.

~/.codex/config.toml

[mcp_servers.siteory]url = "https://api.siteory.co/mcp"

Terminal

codex mcp login siteory

Or do both in one step: codex mcp add siteory --url https://api.siteory.co/mcp adds the server and opens the sign-in page right away.

VS Code

Add Siteory to .vscode/mcp.json in a workspace, or to your user configuration with the MCP: Open User Configuration command.

.vscode/mcp.json

{  "servers": {    "siteory": {      "type": "http",      "url": "https://api.siteory.co/mcp"    }  }}

Start siteory from the file or from MCP: List Servers, then allow the sign-in prompt.

Claude on the web and desktop

Open Customize, then Connectors, and choose Add custom connector. Paste the server URL, click Connect, and sign in. On Team and Enterprise plans, an owner adds the connector in Organization settings first.

Server URL

https://api.siteory.co/mcp

ChatGPT

Add a custom MCP server in ChatGPT developer mode, on plans that include it. Paste the server URL, choose OAuth, and sign in.

Server URL

https://api.siteory.co/mcp

Gemini CLI

Add the server for your user account.

Terminal

gemini mcp add --transport http -s user siteory https://api.siteory.co/mcp

Then run /mcp auth siteory in Gemini CLI to sign in.

Antigravity

Add Siteory to ~/.gemini/config/mcp_config.json, or to .agents/mcp_config.json in a workspace. Antigravity reads remote servers from serverUrl.

~/.gemini/config/mcp_config.json

{  "mcpServers": {    "siteory": {      "serverUrl": "https://api.siteory.co/mcp"    }  }}

Antigravity opens the sign-in page the first time it connects.

OpenCode

Add Siteory to opencode.json, then sign in from a terminal.

opencode.json

{  "$schema": "https://opencode.ai/config.json",  "mcp": {    "siteory": {      "type": "remote",      "url": "https://api.siteory.co/mcp",      "enabled": true    }  }}

Terminal

opencode mcp auth siteory

Windsurf

Add Siteory to ~/.codeium/windsurf/mcp_config.json.

~/.codeium/windsurf/mcp_config.json

{  "mcpServers": {    "siteory": {      "serverUrl": "https://api.siteory.co/mcp"    }  }}

Then refresh the MCP servers in Windsurf and sign in when it asks.

Any other client

Clients that only start local (stdio) servers can reach Siteory through mcp-remote, which runs the browser sign-in for them.

Client config

{  "mcpServers": {    "siteory": {      "command": "npx",      "args": [        "-y",        "mcp-remote",        "https://api.siteory.co/mcp"      ]    }  }}

Sign-in and permissions

Siteory MCP uses OAuth, the same browser sign-in many apps use. Your client never sees your password.

  1. 01

    The first time your agent calls Siteory, your client opens a Siteory page in your browser. Sign in with Google or email.

  2. 02

    Check the app name, whether Siteory could verify it, and where it sends you back. Approve only a connection you started.

  3. 03

    The app gets a token that works for Siteory MCP only. It cannot sign in to your account on the website.

An approved app can

Run Siteory audits, read results and reports, and use your audit credits when you ask.

Connecting needs a paid plan, Single or Pro. On a free account, the sign-in page shows the plans instead of Approve.

Siteory receives the URLs and audit IDs your agent sends. It never reads your repository or your editor.

Connections renew on their own while you use them. After 30 days without use, sign in again. If the account no longer has a paid plan, the app stops working when its access next renews.

To disconnect an app, open Settings, then Connected apps, and choose Revoke. An MCP app stops working right away. The Siteory CLI stops within an hour, when it next confirms your plan.

Tools

Each tool returns compact JSON or Markdown. Read-only tools never start an audit or use a credit.

quick_audit
Quick preview of a public URL: SEO, AI visibility, and a passive security snapshot. Does not use a credit. Up to 3 an hour.
No credit
start_audit
Full audit of a URL you own. Choose SEO, AI visibility, security, or all three. Uses your free full audit, then one credit. An identical request within 10 minutes returns the same audit and is not charged twice.
Uses a credit
get_audit
Progress, scores, and issue counts for one audit. Can wait up to 25 seconds for it to finish.
Read only
list_audits
Your recent audits with their scores and issue counts.
Read only
list_issues
Findings in priority order. Filter by lens or minimum severity, and page through long lists.
Read only
get_issue
One finding with its evidence, affected URLs, references, and a fix prompt your agent can act on.
Read only
get_report
The prioritized, summary, SEO, AI visibility, or security report as Markdown.
Read only
rescan
Runs the same audit again on the same URL and lenses after a fix. Uses a credit, like start_audit. An identical request within 10 minutes returns the same audit and is not charged twice.
Uses a credit
compare_audits
Compares two completed audits and lists the findings that were resolved, introduced, and still open.
Read only
get_account
Your plan, remaining credits, and whether your free full audit is still available.
Read only

Agent skills and the local CLI

Ten skills teach your agent to run Siteory checks and fix the findings in common frameworks. Each one bundles a small CLI that runs on your computer: the hosted audit's SEO, AI visibility and passive security checks, plus local-only lenses for performance, accessibility, structured data, links, share previews and launch readiness. The checks work on localhost. Like Siteory MCP, the skills and the CLI need a paid plan.

Skills and MCP together

Installs the skills for every agent it finds on this computer and adds the MCP server. Add --skills-only to skip the server, or --dry-run to see the changes first.

Terminal

npx siteory setup

Skills CLI

Installs the skills with the open skills CLI, for any agent it supports.

Terminal

npx skills add cyb3r-vishal/siteory-skills

Claude Code plugin

Run these inside Claude Code. The plugin includes the MCP server, so skip claude mcp add or the tools appear twice.

Claude Code

/plugin marketplace add cyb3r-vishal/siteory-skills/plugin install siteory@siteory
siteory
The full loop with the MCP tools: audit, fix, re-scan, and compare.
siteory-seo
Titles, descriptions, headings, canonicals, sitemaps, and robots rules, with fixes by framework.
siteory-ai-visibility
AI crawler access, llms.txt, server-rendered content, and JSON-LD.
siteory-security
Security headers by platform, a safe CSP rollout, and secrets in client code.
siteory-performance
Page weight, render-blocking scripts, image size and format, layout shift, compression, and caching.
siteory-accessibility
Link and button names, form labels, ARIA misuse, zoom blocking, and duplicate ids.
siteory-structured-data
JSON-LD for rich results: Product, Article, FAQ, Breadcrumb, Organization, and more.
siteory-links
Redirecting and broken links, mixed content, leftover localhost or preview URLs, and dead anchors.
siteory-social
Share previews: Open Graph and Twitter tags, preview image size, and reachability.
siteory-launch
One pre-deploy gate: dev builds, soft 404s, template leftovers, broken assets, and every lens at once.

Sign in for the local checker

The CLI confirms your paid plan before it runs a check, so sign in once on each computer. The command opens your browser, where you approve the Siteory CLI. The checks still run on your computer and send nothing about your site to Siteory. Sign-in needs a browser on the same computer, so the CLI does not run in CI.

Terminal

npx siteory login

npx siteory whoami shows the account and plan, and npx siteory logout signs out. Without a sign-in or a paid plan, checks stop with exit code 5.

Check a local build

Point the CLI at a running app. A production build gives the truest results (for example next build && next start), because dev servers add scripts that change what crawlers see. Add --prod https://example.com so checks that compare against your real domain, such as canonical tags, run too.

Terminal

npx -y siteory check http://localhost:3000 --lens seo --json # Every lens before you ship (performance, accessibility, schema, links, social, launch)npx -y siteory check http://localhost:3000 --lens all --prod https://example.com --fail-on high

After a fix, check that one finding again with the same flags. Exit code 0 means it is gone, as long as the run before the fix reported it.

Terminal

npx -y siteory check http://localhost:3000 --only seo_missing_h1 --fail-on info
exit 0
Nothing at or above the --fail-on severity
exit 1
A finding met the --fail-on severity
exit 2
Usage error, such as an unknown flag
exit 3
The start page could not be read
exit 4
Internal error
exit 5
Not signed in, or the account has no paid plan

Example prompts

Ask in plain language. Your agent picks the tools.

  • Audit https://example.com with Siteory and fix every high-severity finding. Re-scan when you are done and show me what changed.

  • Run a Siteory quick audit on https://example.com and summarize the three most important issues.

  • Check http://localhost:3000 with the Siteory SEO skill and fix what it finds in this repo.

  • Compare my last two Siteory audits of example.com and list what is still open.

  • How many Siteory credits do I have left?

Troubleshooting

Sign in again from your client: run /mcp and choose Authenticate in Claude Code, run codex mcp login siteory in Codex, or use the sign-in prompt in Cursor and VS Code. Connections renew on their own while you use them. After 30 days without use, or after you revoke the app in Settings, sign in again.

Run claude mcp login siteory --no-browser in a terminal for Claude Code, or codex mcp login siteory --no-browser for Codex. The client prints a sign-in link that you open yourself.

Siteory only lets you approve a connection in the browser your app opened, so a consent page forwarded from another browser cannot be approved. Close the page and start the connection again from your app. This check does not stop someone from sending you a link that starts a connection to their app, so approve only connections you started.

start_audit and rescan use your free full audit first, then one credit each. An identical request within 10 minutes returns the same audit and is not charged twice. Ask your agent to call get_account to see what is left, or buy more on the pricing page. quick_audit does not use a credit, up to 3 an hour.

Hosted audits only reach public URLs. For a dev server or a private network, use the local CLI: npx -y siteory check http://localhost:3000. It runs the checks on your computer after a one-time npx siteory login.

Your client must reach https://api.siteory.co, and your browser must reach https://siteory.co. Ask your network team to allow both. If sign-in works but tool calls fail, check that the proxy passes the Authorization header through.

You added the server with claude mcp add and also installed the Siteory plugin, which includes the same server. Keep one: run claude mcp remove siteory, or uninstall the plugin.

Yes. Siteory MCP, the agent skills, and the local CLI come with the Single ($9) and Pro plans. Free accounts can use the preview and the free full audit on the website. On a free account, the sign-in page shows the plans instead of Approve, and the CLI stops with exit code 5.

Your next fix is one scan away.

Run a free audit and get a ranked list of what to fix first, with the evidence and a prompt for every issue.

  • First full audit free
  • No credit card
  • Preview in about a minute

Not ready? Browse a sample report