SEO Crawler MCP: Connect Your Crawls to ChatGPT, Claude and Perplexity
An SEO crawler MCP server lets an AI assistant read your crawl data and answer questions about it in plain words. Here is what MCP is, what a connected crawler is good for, and how to connect Neova to ChatGPT, Claude or Perplexity.
What MCP Is
MCP stands for Model Context Protocol, an open standard Anthropic introduced in November 2024 for connecting AI assistants to outside tools and data. Instead of a separate integration for each assistant, a tool builds one MCP server, and any assistant that speaks the protocol can use it.
There are two sides:
- The client is the assistant app: ChatGPT, Claude, Perplexity, a code editor, a command line tool.
- The server is the tool. It tells the client which tools it offers, what each one takes and what it returns.
When you ask a question, the model decides which tools to call, the server answers with data, and the model writes the reply from that data. A server can run on your own computer or on the web, where it is reached over HTTP and usually asks you to sign in with OAuth. The full specification is at modelcontextprotocol.io. MCP is not WebMCP, a proposal that lets agents use a website's own forms; MCP vs WebMCP for non-developers on SEO ProCheck sets the two side by side, and Neova checks WebMCP tools with WebMCP Tool Schema Is Invalid.
What an SEO Crawler MCP Server Is Good For
A crawl holds a lot of data: every URL, its status, its links, and the flags of hundreds of checks. Reading it usually means filters, tabs and exports. With an MCP server, you ask instead:
- Triage: what to fix first on the latest crawl, and on which page groups.
- Explaining: what a check means, in words a client or a developer will follow.
- Writing: a summary for a client email, or ticket text for a developer, built from the real flags.
- Following up: what changed since the last crawl, and which tickets are still open.
- Debugging: why a crawl stopped short or found fewer pages than expected.
The assistant is only as reliable as what it reads, so two habits help: ask it to name the crawl and the date behind each number, and open the flag itself before you act on anything important.
SEO Tools With MCP Servers
Many crawlers and SEO suites offer an MCP server. From each company's own documentation:
| Tool | What it covers | Assistants named |
|---|---|---|
| Screaming Frog | Run crawls, work with data | Claude, LM Studio, others |
| Sitebulb | Finished audits, ready-made Skills | Claude, ChatGPT |
| Lumar | Tasks, crawls, segments | Claude, ChatGPT, Cursor |
| JetOctopus | Start, pause, resume, cancel crawls | Claude, ChatGPT, Cursor, Gemini CLI, n8n |
| Semrush | SEO data | ChatGPT, Claude, Gemini, Perplexity, Cursor, Copilot |
Screaming Frog added its server in version 24.0. Sitebulb's works with Desktop and Cloud, with Skills such as What To Fix First, What Changed and Dev Handoff. JetOctopus offers its server on all plans, and Ahrefs runs one for its SEO data too. The comparison page has the details for each, with sources.
Neova in ChatGPT, Claude and Perplexity
Neova gives assistants two ways in: a local MCP server in the desktop app, and a connector on Neova Online. Both answer from real crawls, and every answer names its crawl and the date the pages were fetched.
| Local server | Online connector | |
|---|---|---|
| Runs on | ||
| Reads crawls from | This computer | Neova Online account |
| Tools | 16 | 12 |
| Start or stop crawls | ✓ | – |
| Delete crawls | ✓ | – |
| Change settings | ✓ | – |
| Sign-in | None | OAuth 2.1 |
| Assistants | Claude Desktop, Claude Code, Cursor, others | ChatGPT, Claude, Perplexity |
The local server adds Neova to Claude Desktop in one click and has copyable setups for Claude Code, Codex, Cursor, VS Code, Gemini CLI, LM Studio, Windsurf and ChatGPT. Every start or delete of a crawl asks your yes first. The online connector only reads. Neova Online comes with the Desktop + Online, Agency and Enterprise plans; see pricing.

How to Connect Neova to Your Assistant
The address is the same everywhere: https://app.neova.app/mcp. The first time you ask Neova something, the assistant shows a sign-in: sign in to Neova Online, read what Neova will share, and choose Allow.
- ChatGPT: open chatgpt.com/plugins, the plus button, then Add custom MCP server. Name it Neova, paste the address, and choose OAuth for authentication.
- Claude: open Customize, Connectors, Add, then Add custom connector. Name it Neova, paste the address, and keep the detected settings.
- Claude Code: run
claude mcp add --transport http neova https://app.neova.app/mcp, then/mcpto sign in. - Perplexity: open Plugins in the sidebar, Add custom, Add MCP connector, Remote. Name it Neova, paste the address, choose Streamable HTTP and OAuth 2.0 (detected), then open the connector card to sign in. In an organization, an administrator adds it under Organization settings, Connectors.
The assistant keeps the connection for 90 days after its last use. You can disconnect at any time from the assistant's connector or plugin page, and Neova forgets it. The Neova in Your AI Assistant help page has the same steps.
What the Assistant Sees, and What It Cannot Do
Neova shares data from your account and nothing else:
- The sites and dates of your crawls.
- For crawled pages: the URL, status, title, description, canonical, robots line, headings, size and inlinks.
- The flags of the checks with their evidence, each check's acceptance line and source.
- Tickets, alert events, page groups, the checks catalogue, your crawl settings and a crawl's diagnosis.
- Your account's email, so the assistant can tell two connected accounts apart.
Text from pages is capped at 500 characters a field, and page bodies are never shared.
The connector cannot start, stop or delete a crawl, change a setting, or read anything outside your account. Those happen in Neova on your computer or on Neova Online.
Questions Worth Asking
Name the site, and the assistant does the rest. To show what comes back, here is what our own full crawl of seoprocheck.com, October 17, 2026, holds for a few questions:
| Question | Tool called | Answer |
|---|---|---|
| How complete was the latest crawl? | crawl_summary | 99.98% |
| How many URLs, and how long? | crawl_summary | 10,222 in 61 minutes |
| Which titles are over 60 characters? | findings | 557 pages |
| Which meta descriptions are over 155? | findings | 582 pages |
| Which titles are duplicated? | findings | 56 pages |
| Which images have no alt text? | findings | 2 images |
Each answer links back to its check: Title Over 60 Characters, Meta Description Over 155 Characters and Duplicate Title Across HTML Pages. Other good questions:
- "Which internal links point to broken pages?" The answer comes from Internal URL Returns 4xx, 5xx or No Response.
- "Which page groups have redirect chains, and how many URLs each?" See Internal URL Is in a Redirect Chain.
- "Are there indexable pages no link reaches?" See Orphan URL Returns 200 and Is Indexable.
- "Is any AI search crawler blocked on key pages?" See Request With an AI Search Crawler User Agent Is Blocked.
- "Write a short summary of the Must Fix flags for my client, with one example URL each."
- "Why did yesterday's crawl find fewer pages than last week's?"
Every check page says when it flags and how to fix it, so you can send the assistant's answer and the source together.
For Developers: the Endpoint and the REST API
Any MCP client that follows the MCP authorization spec connects the same way:
| Item | Value |
|---|---|
| Endpoint | https://app.neova.app/mcp |
| Transport | Streamable HTTP, JSON |
| Authorization | OAuth 2.1 with PKCE |
| Client registration | Dynamic, or metadata documents |
| Metadata | /.well-known/oauth-authorization-server |
| Kept after last use | 90 days |
| Page text per field | 500 characters max |
The protected resource metadata is at /.well-known/oauth-protected-resource on app.neova.app. Claude Desktop and Claude Code on the computer that runs Neova can use the local server described on the Help page instead.
For scripts, spreadsheets and other tools, the REST API at https://app.neova.app/api/v1 reads the same account: portfolios, properties, crawls, flags, URLs, tickets, schedules and alerts. A Read Only key reads; a Read-Write key can also start a crawl in Neova Online and create, update and close tickets. It answers 60 requests a minute per key, and /api/v1/openapi.json describes every endpoint in OpenAPI 3.1.
Questions
What is an MCP server for SEO?
A server that follows the Model Context Protocol and gives an AI assistant access to SEO data, such as crawls, flags or rankings. You ask a question in plain words, the assistant calls the server's tools, and it answers from the data they return.
Can ChatGPT, Claude or Perplexity read my Neova crawls?
Yes. Add Neova as a custom connector with the address https://app.neova.app/mcp, sign in to Neova Online and choose Allow. The steps for each assistant are on the connectors help page.
Can the assistant change my crawls or settings?
No. Neova's connector reads. It cannot start, stop or delete a crawl, change a setting, or read anything outside your account.
Does Neova send my page content to the assistant?
Not the pages themselves. It shares page fields such as the title, description, canonical and headings, capped at 500 characters a field, plus flags, tickets and settings. Page bodies are never shared.
Do I need Neova Online to connect an assistant?
Yes. The connector reads the crawls stored in your Neova Online account. Neova Online comes with the Desktop + Online, Agency and Enterprise plans; see pricing.