Agent access · MCP

Connect your coding agent to MaxiMarket.

MaxiMarket is a remote MCP server for technical SEO and AI-readiness audits. Your coding agent reads the findings and Fix Briefs, fixes the issues in your own repository, and MaxiMarket checks the live result. Claude Code connects in three steps with your MaxiMarket account. Other MCP clients can use a personal API token.

You need a free MaxiMarket account with a confirmed email address.

Server URL
https://maximarket.info/api/mcp
Transport
Model Context Protocol over HTTP
Sign-in
OAuth in your browser (recommended), or a personal API token
Tools
10: three for page audits, seven for Website SEO Audit (Early Access)
Your code
Stays in your environment; MaxiMarket never receives it

Setup

Connect Claude Code

Choose how Claude Code signs in. Most people should use OAuth: there is no token to copy or keep secret.

Claude Code (OAuth) Recommended

  1. Add MaxiMarket

    Run this command in your terminal, not in the Claude Code chat. It adds MaxiMarket for your user, in every project.

    Terminal
    claude mcp add --scope user --transport http maximarket-oauth https://maximarket.info/api/mcp

    The command stores only the server address. There is no password or token in it.

  2. Connect your account

    1. Start Claude Code, or restart it if it is already open.
    2. Type /mcp in Claude Code.
    3. Select maximarket-oauth, then choose Authenticate.
    4. Your browser opens. Sign in to MaxiMarket.
    5. Check the permissions MaxiMarket shows, then choose Allow.

    Sign-in starts in Claude Code; this website has no Connect button for it. If the browser does not open, open the link Claude Code shows. You can also start sign-in from your terminal with claude mcp login maximarket-oauth.

    The first connection can only read: your page audits and, with Early Access, your Sites, crawls, findings and fix checks. It cannot start anything until you allow that separately.

  3. Try your first prompt

    Type this in Claude Code:

    Prompt
    Use the MaxiMarket MCP connection to list my websites and show the latest SEO audit findings for one of them.

    Sites, crawls and findings need MaxiMarket Early Access on your account.

    To start something, such as a crawl, try:

    Prompt that starts a crawl
    Use MaxiMarket to preview a new crawl of one of my websites with a dry run. Show me the settings and the remaining quota, and start the crawl only after I confirm.

    Starting a crawl needs the crawl:create permission and uses your crawl quota. The first time, the call fails and Claude Code says it needs additional permissions. Run /mcp, authenticate maximarket-oauth again, and allow “Start Website SEO Audit crawls” in your browser. Fix checks work the same way with fixcheck:create. Claude Code also asks before it uses a tool, unless you have allowed that tool.

Personal API token Advanced

Use a token for scripts, CI, MCP clients without browser sign-in, or when your agent should start single-page audits (run_audit). Create one under Agent access in your dashboard and choose its permissions. It is shown once.

Keep the token out of commands, prompts and repositories. Store it in an environment variable, and point your configuration at the variable. Claude Code reads ${MAXIMARKET_API_TOKEN} from your environment when it connects.

.mcp.json
{
  "mcpServers": {
    "maximarket": {
      "type": "http",
      "url": "https://maximarket.info/api/mcp",
      "headers": {
        "Authorization": "Bearer ${MAXIMARKET_API_TOKEN}"
      }
    }
  }
}

Never paste a token into an AI chat. If a token is exposed, revoke it under Agent access and create a new one.

Other MCP clients

Server URL
https://maximarket.info/api/mcp
Transport
Streamable HTTP
Header
Authorization: Bearer <your token>

Clients that support OAuth sign-in for remote MCP servers can connect without a token, as Claude Code does.

Access and permissions

Three separate checks on every request.

MaxiMarket permissions and how Claude Code (OAuth) and personal API tokens get them
Permission Lets your agent Claude Code (OAuth) API token
audit:read Read your page audits and Fix Briefs. When you connect If you choose it
sites:read Read your Sites, crawls, findings, Agent Briefs and fix check results. When you connect If you choose it
crawl:create Start a Website SEO Audit crawl. Only after you allow it If you choose it
fixcheck:create Start a fix check. Only after you allow it If you choose it
audit:create Start a single-page audit. Not available If you choose it

You can see and remove connected apps and tokens under Agent access at any time. Removing one stops it at its next request.

MCP tools

10 tools, grouped by what they do.

A page audit checks one URL and is available to every account. Website SEO Audit crawls a whole site; its tools need MaxiMarket Early Access. Each tool also needs the permission shown.

Page audits Every account

A page audit checks one public URL. Your agent can read your audits and their Fix Briefs.

  • get_audit(audit_id)audit:read

    Status, Technical Score and findings of a page audit.

  • get_fix_brief(audit_id)audit:read

    The structured Fix Brief and its agent contract.

  • run_audit(url)audit:createAPI token only

    Starts an audit of one public page.

Website SEO Audit: read Early Access

Website SEO Audit crawls up to 500 pages of a saved site and groups findings across pages.

  • list_sites()sites:read

    Your Sites and their latest crawl.

  • get_crawl_status(crawl_id)sites:read

    A crawl’s progress, counts and Audit Standard.

  • get_findings(crawl_id)sites:read

    A crawl’s findings by rule.

  • get_agent_brief(crawl_id, rule_ids)sites:read

    An Agent Brief for up to 20 findings.

  • get_fix_check_results(fix_check_id)sites:read

    What a fix check found: fixed, still failing, not verified.

Website SEO Audit: start Early Access

These start work on MaxiMarket’s servers and use your quota. Each has a preview (dry_run) that starts nothing.

  • start_crawl(site_id)crawl:create

    Starts a crawl of one of your Sites. Uses your crawl quota.

  • start_fix_check(crawl_id, rule_ids, mode)fixcheck:create

    Checks the pages behind chosen findings again. Uses your fix check quota.

An Agent Brief explains chosen findings of a crawl to a coding agent. A fix check (Verify fixes) audits the affected pages again and shows what is fixed.

Troubleshooting

If something does not work

maximarket-oauth does not appear in /mcp

Run claude mcp list in your terminal. If MaxiMarket is missing, run the command from step 1 again, then restart Claude Code. With --scope user it is available in every project.

Sign-in did not finish

Run /mcp, select maximarket-oauth and authenticate again. If the browser does not open, open the link Claude Code shows. Sign in with an account whose email address is confirmed.

A tool needs additional permissions

Starting a crawl or a fix check needs a permission the first connection does not include. Run /mcp, authenticate maximarket-oauth again, and allow the permission in your browser. If you set oauth.scopes for this server in your Claude Code settings, add crawl:create or fixcheck:create there first.

“Connected apps cannot use this operation” means the tool is for API tokens only, such as run_audit. Start the audit on maximarket.info, or use a personal API token.

“Website SEO Audit needs MaxiMarket Early Access”

The tool answers FEATURE_NOT_ENABLED: Early Access is not active for your account. See the Sites page of your dashboard. Page audit tools keep working.

“The access to MaxiMarket was revoked”

Claude Code was removed under Agent access, or its sign-in ended. Run /mcp, select maximarket-oauth and authenticate again.

New or changed tools do not show up

Claude Code reads MaxiMarket’s tool list when it connects. Reconnect maximarket-oauth from /mcp, or start a new Claude Code session.

The loop

Audit, fix, deploy, verify.

  1. MaxiMarket

    Audits the live page

    Your agent asks for an audit of a public URL, checked against the published Audit Standard.

  2. MaxiMarket

    Returns a Fix Brief

    Each technical SEO, structured data or AI crawler finding, with evidence and the expected state.

  3. You or your agent

    Implements the fixes

    In your repository and your own environment, changing only what the findings require.

  4. You or your CI

    Deploy

    You decide what ships and when.

  5. MaxiMarket

    Re-audits the page

    A finding is resolved only when a new audit of the same URL no longer reports it.

Your repository stays with you

MaxiMarket evaluates the website. It never sees your code.

The coding agent runs where you run it, with the AI provider or local model you choose. MaxiMarket stays the independent evaluator of the deployed result.

MaxiMarket receives
Page URLs to audit and the IDs of audits to read.
MaxiMarket returns
Audit results and structured Fix Briefs.
MaxiMarket never receives
Your source code, your repository or access to your deployment. It never edits or deploys your code.

The Fix Brief

Everything an agent needs, nothing it should not touch.

Evidence
What the audit detected, the audited and final URL and the HTTP status.
Expected state
What the page should declare or return, from the published rule.
How to fix
The change that resolves the finding, and why it matters.
Scope
Whether the finding is page-level or comes from site-wide code.
Verification
The re-audit target and the exact pass condition.
Sources
The documentation behind the rule, such as Google Search Central or Schema.org.

The agent contract in every brief: inspect the whole repository if needed, change only what resolves the findings (shared code only where a finding originates), no unrelated cleanup, refactoring, content or design changes, then re-audit the original URL.

Ready to connect your agent?

Add MaxiMarket to Claude Code, sign in, and ask your agent to read your latest findings.