Published · Last updated
Ask for the report. Launch the campaign. Phished now runs from your AI assistant
Phished is the first security awareness training platform you can run by asking Claude. Anything you would normally click through the platform to do, from pulling a board-level report to launching a phishing campaign, can now be asked for in plain language. It is not tied to Claude: the same connection works in ChatGPT, Gemini or any other MCP client.
The problem
Security teams do not lack features. They lack time in the tool. A platform can hold every report an admin could want and still go unread, because pulling the report means remembering which screen it lives on, going there, filtering it and exporting it, and that is a twenty-minute errand on a day that has none to spare. The work the platform does well is the work nobody has time to ask it for.
Why the platform should come to you
MCP is an open standard that lets AI assistants use a product’s tools directly. We built the Phished MCP server on it so the platform arrives where an admin already works, instead of asking them to come to us. There is no new dashboard to learn and no query language to write. You ask for what you want in the words you would use with a colleague.
Over 90 tools are exposed through it, which is the platform rather than a summary of it. Ask for the numbers your board meeting needs and they come back as an answer. Ask for a simulation campaign to go out to a department and it goes out.
It works strictly within your own permissions. This is the part we would not compromise on. An assistant that can see more than you can is not a convenience, it is a privilege escalation with a friendly interface, so nothing is exposed through MCP that the person asking could not already see or do in the platform themselves. Your access is the ceiling, in both directions.
Adding it takes seconds, and because it is an open standard it is not a bet on one vendor. If your organisation standardises on a different assistant next year, the connection follows you.
Frequently asked questions
Which AI assistants does it work with?
Claude, ChatGPT, Gemini and any other MCP client. MCP is an open standard, so support is not limited to a list we maintain. The Phished connector for Claude is the quickest way in if you are already using Claude.
What can I actually ask it to do?
Anything you would click through the platform for. Over 90 tools are available, covering reporting and campaigns among the rest, so a board-level report and a new phishing simulation are both a sentence rather than a session.
What are the security considerations for the Phished MCP server?
The permission boundary is the one that matters, and it is absolute: the MCP server works strictly within your own permissions. Nothing is exposed through it that you could not already see or do yourself in the platform, so connecting an assistant cannot widen what your account reaches.
How do I connect it?
It takes seconds, and the same connection works in Claude and in any other MCP client.
Is this only useful for admins?
Partners get the same connection, within the same permission boundary, so a partner can ask about the accounts they manage in the same way an admin asks about their own organisation.
Want to try it on your own data?
Request a demo and we’ll connect an assistant to a Phished environment and ask it something real.