Cursor Agent
Operated by Anysphere
Quick Facts
- User-Agent:
- Cursor
- Category:
- AI & LLM Bots
- Operator:
- Anysphere
- Safety:
- Safe
- Blocking Impact:
- Low - No SEO ranking impact
- SEO Impact Score:
- 2/10
- Official docs:
- None published by Anysphere
Reported, not operator documented
Anysphere has not published a crawler page that names Cursor. It is listed here on our own evidence. Verified by AI Crawler Check: 16 of the 5,583 valid robots.txt files in the Tranco top 10,000 carry a rule for Cursor, so site owners treat it as a real crawler name (how we measure this). That proves the spelling is in use, not who sends it. Reported Cursor agent browser fetches. No vendor UA published.
What that means for you: the Cursor rule on this page is the best available spelling, but Anysphere has made no statement of purpose, crawl rate or robots.txt compliance. Treat a Cursor request in your logs as a clue, not proof of who sent it, and add a firewall rule if that traffic matters to you. If the operator publishes documentation, this profile moves to Official docs.
What is Cursor Agent?
Cursor Agent is an on-demand AI fetcher operated by Anysphere. It retrieves a page only when a user explicitly asks an AI assistant or agent to read, summarize, or act on a specific URL. Because it is triggered by real user intent rather than bulk crawling, blocking it mainly prevents users from pulling your content into Anysphere's assistant.
Cursor Agent is an on-demand AI fetcher operated by Anysphere. It retrieves a page only when a user explicitly asks an AI assistant or agent to read, summarize, or act on a specific URL. Because it is triggered by real user intent rather than bulk crawling, blocking it mainly prevents users from pulling your content into Anysphere's assistant.
Cursor Agent uses the user-agent token Cursor. You can control it via robots.txt, meta tags (noai), or the emerging llms.txt standard. Robots.txt is voluntary; for hard enforcement, combine it with server-level IP blocking.
What happens if you block Cursor Agent?
User-agent / Disallow: / without any SEO penalty.Want to see whether Cursor Agent actually visits you? Read how do I monitor AI bot traffic to my website?
How to block Cursor Agent with robots.txt
<code>User-agent: Cursor</code> - Matching is case-insensitive. Robots.txt is fetched from the root of each subdomain separately.
Is Cursor Agent safe to allow?
What does Cursor Agent do?
Understanding Cursor Agent's purpose helps you decide whether to allow or block it.
- Fetches a page only when a user asks an AI assistant to read it
- Summarizes or extracts the URL the user provided
- Triggered by real user intent, not bulk crawling
- Low traffic volume compared with training crawlers
- Blocking it prevents users from pulling your content into the assistant
Cursor Agent in context
Cursor Agent is a user-triggered agent operated by Anysphere. In plain terms, it fetches a page when a person asks an assistant or app to open it. Our database files it under AI and LLM crawlers and ranks it Tier 3, mid-priority crawlers with a clear purpose but a smaller audience. 15 of the 104 AI and LLM crawlers in our database share this tier.
Its safety rating is safe and its blocking impact is low. So the decision about Cursor Agent rests on what you want from Anysphere, not on its behaviour. How we rate safety and impact.
How site owners treat Cursor Agent
In our own scan of the Tranco top 10,000 (5,583 valid robots.txt files), 16 files (0.3%) name Cursor in a User-agent line. That shows site owners treat the token as a real crawler worth a rule. It does not, on its own, prove who operates it.
Does it follow robots.txt?
Cursor Agent's operator does not clearly document how it treats robots.txt. Write the Cursor Agent rule anyway, then confirm in your logs that its requests stop.
Anysphere crawlers and close alternatives
Anysphere runs only this one crawler in our database, so a single User-agent: Cursor group controls everything Anysphere fetches through it.
Other AI and LLM crawlers that do the same job (fetches a page when a person asks an assistant or app to open it) include ChatGPT Agent from OpenAI, ChatGPT Operator from OpenAI, ChatGPT-User from OpenAI, ChatGPT-User/2.0 from OpenAI, Claude-SearchBot from Anthropic, Claude-User from Anthropic. Deciding about Cursor Agent is usually a decision about the whole group, so check how your file treats these too.
We first recorded Cursor on September 22, 2026. We could not find a page where Anysphere names it, so treat the token as reported rather than documented.
Which rule Cursor Agent actually follows
Cursor Agent reads exactly one group in robots.txt, the most specific one that names Cursor, and never combines rules across groups. This short file shows the trap:
User-agent: *
Disallow: /private/
User-agent: Cursor
Disallow: /drafts/- Find the group. Cursor Agent looks for a group naming
Cursor. One exists, so it ignoresUser-agent: *completely. - Apply only that group. It is refused
/drafts/and nothing else. - The surprise.
/private/is open to Cursor Agent, because that rule lives in the*group it no longer reads. To keep it out of both, repeatDisallow: /private/inside its own group.
Test yourself: does a rule under User-agent: * apply to Cursor Agent if the file also has a User-agent: Cursor group?
No. Once a group names it, Cursor Agent reads only that group. Put every rule you want it to follow inside its own group.
Related: how AI engines choose their sources
Test yourself: will blocking Cursor Agent remove your pages from AI answers?
Often, yes. Cursor Agent is a user-triggered agent, so pages it cannot fetch can drop out of the answers or results it supports.
Test yourself: is User-agent: cursor the same as User-agent: Cursor?
Yes. RFC 9309 makes user-agent matching case-insensitive. Paths are a different story: Disallow: /Blog/ does not block /blog/.
Should you allow Cursor Agent? Decide by situation
| If you are... | Recommended setting |
|---|---|
| You publish content and want AI answers to cite it | Allow it. Cursor Agent fetches pages for answers, so blocking it can remove you from them. |
| You license content and want to stop AI training | Blocking it does not stop training, because Cursor Agent is a user-triggered agent. Block the operator's training crawler instead. |
| You run an online shop | Allow it on product and category pages, which are the pages shoppers ask assistants about. |
| Your server is under load from bots | Rate-limit before blocking. It is rated safe in our database. |
Check Cursor Agent on your own site
- Open
/robots.txton your domain and search forCursor. If it is not named, it followsUser-agent: *. - List the
Disallowlines in the group Cursor Agent follows and test your key URLs against them; the longest match wins, andAllowwins a tie. - Look beyond the file. A firewall or bot manager can refuse Cursor Agent even when robots.txt allows it, and that never shows in the file.
- Run our AI bot access checker to see how Cursor Agent and every other crawler we track are treated in one pass, including firewall refusals.
Common mistakes with Cursor Agent
The errors we see most often with Cursor Agent in the files and firewalls we test:
- Assuming a robots.txt rule will stop Cursor Agent. User-triggered fetches act for a person, and some operators say robots.txt may not apply to them. Check the operator's statement above.
- Blocking Cursor Agent as if it were a bulk crawler. It fetches one page because a person asked; blocking it means that person's assistant cannot read your page.
- Counting Cursor Agent visits as crawl volume. Each request maps to a real user question, which makes the paths it requests a useful signal of what people ask about you.
All three hide inside a file that looks correct, so check Cursor Agent from outside after every change.
Spotting Cursor Agent in your logs
robots.txt says what Cursor Agent may do; your access logs show what it actually did.
Count requests
On a server with standard Nginx or Apache logs, this counts matching requests per day (adjust the log path for your host):
grep -iE "Cursor" /var/log/nginx/access.log \
| awk '{print $4}' | cut -d: -f1 | sort | uniq -cAnd this shows which paths it requested most, which is the quickest way to see whether your Disallow lines are being respected:
grep -iE "Cursor" /var/log/nginx/access.log \
| awk '{print $7}' | sort | uniq -c | sort -rn | head -20Check the requests are genuine
Anyone can send the Cursor Agent user-agent, so confirm the source before acting on a match:
Cursor: we hold no published IP list, so use reverse DNS where the operator documents a hostname and treat an unverifiable Cursor Agent request with caution.
Frequently Asked Questions
What user-agent string does Cursor Agent use?
Cursor. Evidence that the spelling is in real use comes from our own robots.txt study: 16 of the 5,583 valid robots.txt files in the Tranco top 10,000 name it (fetched October 20, 2026), not from Anysphere, which has not published crawler documentation naming it. Use it in robots.txt as the best available match, and treat a request carrying it as a clue rather than proof of who sent it.Is Cursor Agent safe?
Will blocking Cursor Agent hurt my SEO?
User-agent / Disallow: / without any SEO penalty.How do I block Cursor Agent in robots.txt?
/robots.txt file:User-agent: Cursor Disallow: /
This instructs Cursor Agent not to crawl any path on your site. To block only specific sections, replace / with the path (e.g.,
Disallow: /blog/).Does Cursor Agent respect robots.txt?
Cursor will stop a client that chooses to honour it; for enforcement, add a firewall or CDN rule as well.How do I verify if Cursor Agent is crawling my site?
Cursor (for example grep -i "Cursor" /var/log/nginx/access.log). Because the operator publishes no IP ranges or reverse-DNS pattern, a match shows that something used the name, not that Anysphere sent it.