Check if a URL is phishing or malware before an agent opens it
How can an AI agent check whether a link is phishing or malware before opening it?
Call GET /v1/safety/url-check with the url. Without visiting it, Octio checks the URL against public threat feeds (URLhaus by abuse.ch and PhishTank) and reads static signals from the URL itself, such as look-alike brand names, punycode hosts and URL shorteners, and returns a verdict: listed_malicious, suspicious or no_known_issues.
Example
Example call: safety-url-check
What is checked
- Threat feeds: URLhaus (abuse.ch) and PhishTank, downloaded in the background. sources lists each matching feed and feeds when it was downloaded.
- Static signals: look-alike brand names, punycode and mixed-script hosts, IP-address hosts, text before an @, URL shorteners and heavily abused TLDs.
The URL is never fetched. no_known_issues is not a guarantee: new phishing pages appear before any feed lists them. The call is not charged while no feed is loaded.
FAQ
Does the check visit the link?
No. It only compares the URL with downloaded feeds and reads the URL’s own text, so nothing on the page can run.
What should an agent do with suspicious?
Treat it as a reason not to open the link without a person’s approval. The signals explain why it was marked.
Updated 2026-11-16.