Claude adds VirusTotal threat intelligence for AI-assisted malware and URL checks
in plain words
Claude, an AI assistant, can now link to VirusTotal, a service that checks files, web addresses and domain names for danger. Security workers can ask Claude to explain a report about a file or link. The tool does not decide whether something is safe.
Full summary
Claude users can connect VirusTotal threat intelligence to their investigations. The connection brings file, URL, domain and IP Internet Protocol The internet's basic rules for sending data, using numeric addresses for every connected device. Full definition of IP reports into an AI workflow, so analysts can review detection findings, analysis dates, coverage and source links. The integration uses the Model Context Protocol, known as MCP, to link Claude with VirusTotal AI, called VTAI. Users sign in through browser OAuth and do not need a personal VirusTotal API key. Organization rules may require an owner to approve the connector before employees can use it.
The main benefit for security teams is evidence review. An analyst can provide a file hash or a suspicious address and ask Claude to explain the report. The source says missing reports and zero detections do not establish safety, so analysts still need to consider report age and coverage.
Report lookups are separate from submissions. Only an explicit user request authorizes file uploads, URL submissions or domain and IP reanalysis. Before uploading a file, the service checks its verified hash.
An optional plugin for Claude Code adds a threat intelligence skill and a local file hook, according to VirusTotal’s documentation. Claude chat does not run this local hook, and the remote connector cannot read local paths. Hosted Claude chat report workflows remain unverified in testing.
The official connection guide lists free access at 60 queries per minute and 1,000 per UTC day, shared across account connections. Standard submissions are shared with VirusTotal’s community and security partners, so private documents, internal code, credentials and personal data require permission before disclosure. URL queries expose the complete address. Users can revoke access through their connections, but removing the plugin does not erase prior submissions or receipts held by the service.