Your attack surface, inside the AI tools you already use
Connect Claude, ChatGPT, Cursor, Claude Code or VS Code to FortWatch. Ask about your assets, scans and issues, run a scan, or triage, as you, with your role. Anything that hides an issue or weakens monitoring waits for your click in FortWatch.
Claude · ChatGPT · Cursor · Claude Code · VS Code
Your AI tool
Connected to FortWatch · mcp.fortwatch.ai
What are my critical issues in FortWatch?
One critical: an exposed .env file on staging.example.com. Two high: a Webmin admin panel open on port 10000, and a certificate that expires in 6 days.
Rescan staging, then mark the .env issue fixed
The scan of staging.example.com has started. Marking the .env issue resolved needs your confirmation in FortWatch. Nothing has changed yet.
Confirm in FortWatch · 15 minConnected in a minute
One address works in every AI tool that speaks MCP. You sign in with FortWatch, so there is no key to create, copy or leak.
Add the server
In your AI tool's connectors or MCP settings, add FortWatch with one address. No key to copy, nothing to install.
Sign in and approve
Your AI tool opens FortWatch. Sign in, enter your two-factor code, and approve the areas it may use.
Ask
“What are my critical issues?” “Scan staging now.” The AI tool works with your live FortWatch data, as you.
https://mcp.fortwatch.ai/mcp17 tools: reads and real actions
The same tools FortWatch AI uses inside the dashboard, with the same roles, plan limits and audit log. A viewer’s AI tool gets the reads; every other role gets both.
11 reads
- Projects, assets and one asset in detail
- Scans, one scan's results, and scan schedules
- Issues and findings, each in detail with its fix
- A security overview of the whole workspace
6 actions
- Run a scan on an asset
- Create a scan schedule, or change one
- Change an issue's status
- Assign an issue to a teammate
- Dismiss an issue
Built for AI tools that can be tricked
An AI tool that can browse, send email and change your security settings is a target. FortWatch assumes it can be misled, and limits what a misled one could do.
It acts as you
The AI tool works in your workspace with your role and only the areas you approved. A viewer's AI tool reads; it cannot change anything.
Risky changes wait for your click
Dismissing an issue, marking it resolved, or pausing or slowing a schedule returns a link to FortWatch. Nothing changes until you confirm there, within 15 minutes. The AI tool cannot confirm for you.
Text from scanned websites stays out
A scanned site can plant instructions in its page title, banner or DNS records. FortWatch sends the AI tool its own words and plain facts, never what a website wrote, so planted text cannot steer your AI tool.
Every change in the audit log
Each change carries your name and the AI tool's, such as “Anna via AI tool ‘Claude’”, so you always know who did what, and through what.
Two-factor to connect
Connecting an AI tool needs two-factor authentication. Changing your password, or signing out of all other devices, disconnects your AI tools.
Disconnect any time
Your profile lists every connected AI tool. Disconnect one and its access ends at once.
Prefer to stay in FortWatch? FortWatch AI runs the same tools from the dashboard’s chat box. Building your own automation? Use the REST API.
Frequently asked questions
Answers to the most common questions about connecting AI tools to FortWatch.
What is the FortWatch MCP server?+
MCP, the Model Context Protocol, is the open standard AI tools use to work with other products. FortWatch's MCP server at mcp.fortwatch.ai lets Claude, ChatGPT, Cursor, Claude Code and VS Code read your assets, scans, issues and findings, run scans, and update issues and scan schedules, as the person who connected them.
How do I connect my AI tool?+
Add https://mcp.fortwatch.ai/mcp as a connector or MCP server in your AI tool, then sign in to FortWatch when it asks, enter your two-factor code and approve. The setup guide at docs.fortwatch.ai/mcp has the steps for each AI tool.
What can my AI tool do in FortWatch?+
Eleven reads: projects, assets, scans, scan schedules, issues, findings, each in detail, and a security overview. Six actions: run a scan, create or change a scan schedule, change an issue's status, assign an issue, and dismiss an issue. A viewer's AI tool gets the reads only. Adding or deleting assets, team changes and settings stay in FortWatch.
Can an AI tool hide a real issue or switch off my monitoring?+
Not by itself. Dismissing an issue, marking it resolved, and pausing a schedule or making it scan less often answer with a link to FortWatch, and nothing changes until you open it and confirm, within 15 minutes. Only you see that request, and only from your FortWatch session, so the AI tool cannot confirm it for you.
What about prompt injection from scanned websites?+
FortWatch scans websites it does not control, and any of them can write instructions into a page title, a server banner, a header or a DNS record. FortWatch sends AI tools its own words and plain facts (names, severities, statuses, CVEs, hosts, ports, dates and the fix), and leaves out the text the website wrote, so planted instructions never reach your AI tool.
Which plans include the MCP server?+
Every plan. Your AI tool reaches what your plan allows in the dashboard, with the same limits: each connection may make 120 reads and 60 changes a minute.
Can I use an API key instead of signing in?+
Owners and admins can, but signing in is the better way. A key also works on the REST API, where changes ask nobody, while a signed-in AI tool can only use the MCP server and its risky changes wait for your click.
Explore more of FortWatch
Every feature runs on one platform and shares the same data. Jump to another, or see the full set.
Scanning & discovery
Secure your entire stack today
Start scanning in under 5 minutes. No credit card required. 7-day free trial included.