squirrelscan for ChatGPT
Turn on developer mode, add squirrelscan as a custom connector, and run a full website audit without leaving the chat.
ChatGPT doesn't run a CLI or have a terminal, so this one works differently from our other tool pages. Instead of installing anything, you add squirrelscan as a custom connector in developer mode. From there you just ask: audit a URL, and squirrelscan's hosted MCP server does the crawling, the 260+ rule analysis, and the scoring in the cloud, then hands the results back into the conversation.
Worth saying plainly: ChatGPT is the interface here, not the engine. The audit itself runs on squirrelscan's infrastructure with real browser rendering, the same cloud audits you'd trigger from the dashboard or the API. That also means it spends credits, and ChatGPT is expected to show you the estimate before it starts anything.
Setup
Run it hosted
cloud · rendering, scheduling, issue tracker, shareable reports- 1
Turn on developer mode
In ChatGPT on the web, open Settings, then Security and login, and turn on Developer mode. It's available on Plus, Pro, Business, Enterprise, and Edu accounts, and gives custom connectors the full set of MCP tools, both read and write. On a Business, Enterprise, or Edu workspace, an admin has to enable developer mode first under Workspace settings, then Permissions.
- 2
Add squirrelscan as a custom connector
In Settings, open Apps & Connectors, then Create. Give the connector a name, set the MCP server URL to the hosted squirrelscan endpoint, and choose OAuth for authentication. The transport is streamable HTTP, which ChatGPT negotiates for you.
Hosted MCP server setuptexthttps://mcp.squirrelscan.com/mcp - 3
Sign in and pick your organization
Approving the connector runs the same OAuth flow as every other squirrelscan MCP client: you sign in, choose which organization the agent should act for, and approve. Nothing to copy or paste.
- 4
Ask for an audit
Once the connector shows as connected, just ask in plain language. The first call returns a credit estimate instead of starting anything, so ChatGPT shows you the cost and waits for your go-ahead before the audit runs, usually a few minutes for a mid-size site.
How cloud credits worktextAudit https://example.com with squirrelscan and summarize the top issues by severity - 5
Get the report
Ask for a summary for a quick read, or the full detail if you're planning fixes. ChatGPT pulls it straight from the finished report: scores, top issues grouped by rule, and the fix text for each.
Fix your site with an AI agenttextGive me the full report, and tell me what changed since the last audit
Frequently asked questions
- Can ChatGPT audit my website's SEO directly?
- Yes, once squirrelscan is added as a custom connector in developer mode. ChatGPT calls the hosted MCP server, which runs the crawl and the 260+ rule analysis in squirrelscan's cloud and returns the report into the chat.
- Do I need a paid ChatGPT plan to add the connector?
- Developer mode and custom MCP connectors are available on the web for Plus, Pro, Business, Enterprise, and Edu accounts, not the free tier. On a Business, Enterprise, or Edu workspace, an admin enables developer mode before members can add a connector. You also need a squirrelscan account, since cloud audits spend credits.squirrelscan accounts and keys
- Will it spend credits without asking?
- No. The first run_audit call returns a cost estimate instead of starting anything, and ChatGPT is expected to show you that estimate and get a confirmation before the audit runs. Every audit hits the confirm step, since each one costs at least the base credits.
- Does the audit actually run inside ChatGPT?
- No. ChatGPT is the interface. The crawl, browser rendering, and rule analysis all run on squirrelscan's cloud infrastructure, the same engine behind the CLI and dashboard. ChatGPT just calls the tools and shows you the result.
- How do I get ChatGPT to fix the issues it finds?
- Ask for the full report and ChatGPT gives you the exact fixes as text, grouped by rule. ChatGPT can't edit your repo directly, so to apply the changes in code, connect the same squirrelscan MCP server to a coding agent like Claude Code, Cursor, or Codex.MCP client setup for coding agents
Get started
Follow the docs quickstart, or see what a finished audit looks like.