CrawlBit MCP
Four free tools that answer one question: can AI engines find, read and recognise a website?
No account. No API key. No signup. Nothing here runs a language model, so nothing here costs you
or us anything.
"Can ChatGPT read stripe.com?"
"Why does AI never mention my store?"
"Where is my brand missing outside my own site?"
Ask in plain language. Your AI client picks the right tool.
Install
Requires Node.js 18+.
Claude Code
claude mcp add crawlbit -- npx -y crawlbit-mcp
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"crawlbit": {
"command": "npx",
"args": ["-y", "crawlbit-mcp"]
}
}
}
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"crawlbit": {
"command": "npx",
"args": ["-y", "crawlbit-mcp"]
}
}
}
Restart your client. Four tools appear.
| Tool | What it answers |
|---|
crawlbit_crawler_watch | Can GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended read this site? |
crawlbit_entity_check | Does this brand read as a clear entity AI can recognise? |
crawlbit_offpage_gaps | Where is this brand missing outside its own website? |
crawlbit_shopping | Can AI shopping agents understand and recommend these products? |
Try these
Check whether AI crawlers can read example.com
Does example.com read as a clear brand entity to AI?
Where is example.com missing off-page, and which gap should we close first?
Compare example.com and competitor.com on AI crawler access
The tools compose. A useful sequence is: crawler access first (a blocked crawler makes everything
else pointless), then the entity check, then the off-page gaps.
What this does NOT do
Written plainly, because an SEO tool that overstates itself is worse than no tool.
- It does not test whether AI actually cites a brand. That means running real buyer questions
against ChatGPT, Perplexity and Claude, reading the answers, and reporting who gets named
instead of you. It costs real money per run and is not part of this free surface. These four
tools measure the conditions that make a citation possible, which is a smaller claim.
crawler_watch reads robots.txt only. It cannot see blocking done at the CDN, firewall
or rate-limit layer. A site can pass here and still refuse the crawler in practice.entity_check reads what the site publishes about itself. It says nothing about the
brand's reputation across the rest of the web.shopping inspects published structured data. It cannot see a merchant feed submitted
privately to a platform.
If a site scores well on all four and still is not cited, the answer is almost always the same and
it is not technical: too few third-party pages mention it. These tools will tell you that
honestly rather than sell you a fix that does not exist.
CrawlBit also runs a full technical audit and generates llms.txt. Both use a language model,
so both cost money per run, and neither is exposed here.
That is a deliberate line rather than a teaser. A free tool whose bill grows with its popularity
gets rate-limited, degraded or withdrawn the moment it succeeds, and the people who installed it
are the ones who pay for that. Everything in this server costs nothing to run, so nothing here
has to be taken back later. The audit and the llms.txt generator live at
crawlbit.app if you want them.
Rate limit
The API rate-limits per IP. Normal conversational use does not come close, and the server runs on
your machine, so the budget is yours rather than shared. If you do hit it, you get a plain
sentence rather than a stack trace.
Configuration
| Variable | Default | Purpose |
|---|
CRAWLBIT_BASE_URL | https://www.crawlbit.app | Point the server at another instance |
Privacy
The server runs locally and stores nothing. The only data leaving your machine is the URL you
ask about, sent to the CrawlBit API to be analysed. There is no account, so there is nothing to
tie a request to a person.
About
Built by CrawlBit, which measures AI visibility and does the off-page
work that earns citations. The free tools here are the measurement half. The paid work is the
other half, and you are not required to look at it to use this.
MIT licensed. Issues and pull requests welcome.