Security scan: a one-click pen test for your deployed site
Most sites ship with a few security gaps that nobody meant to leave: a missing Content-Security-Policy, a session cookie without the Secure flag, a .env file that got deployed by accident. None of them show up when the app “works,” and all of them are the first things an attacker checks. Today we are adding a security scan to Coaade IDE: a one-click, non-destructive pen test of your own deployed site.
What it checks
The scan looks at the same things a pen tester checks first — the response your site actually sends, its cookies, and a short list of files that should never be public:
Security headers CSP, HSTS, X-Frame-Options, nosniff, Referrer-Policy Cookies Secure, HttpOnly, SameSite Transport http served without a redirect to https CORS reflects any origin, or wildcard with credentials Disclosure Server / X-Powered-By versions Exposed files .env, .git, .npmrc, credentials, backups
Every finding comes with a severity and a specific fix — not “add a CSP” in the abstract, but what the header does and where to set it. A clean scan is a good baseline; it finds the common misconfigurations, not every possible vulnerability.
Non-destructive, and only your own site
This is the important part. The scan sends only plain GET requests that read headers, cookies and a small sample of the response. It sends no attack traffic — no injection, no fuzzing, no brute force, no load — and it changes nothing. And like App Test, it is locked to your own property: it resolves the target to your project's coaade.app deploy or a custom domain you have verified, so it can never be pointed at someone else's site. It is an audit of what you ship, not a weapon.
Run it from the sidebar, or over MCP
In the IDE, the security scan has its own view, opened from the left sidebar: one click runs it against your live site and lists the findings, each with its fix. You can also run it from the AI tools you already use — through Coaade MCP, ask ChatGPT, Claude or Cursor to “run a security scan on my site” and the findings come straight back in the conversation. Past scans are saved, so you can compare before and after a fix.
Want it to actually attack the app? Run the AI pen test
The scan is passive — it reads what your site already sends. For a deeper pass, the same Security & Code review tab has Run AI pen test: an AI agent that opens your deployed site in a real AWS cloud browser and sends crafted (still safe, non-destructive) requests, hunting what a penetration tester and QA engineer look for — authentication and sessions, access control and IDOR, injection, data exposure, CORS, and broken user flows. It reports each finding with evidence and a fix and ends with a 0–100 security score. Like the scan, it only ever tests your own site, uses safe verbs, and confirms weaknesses rather than exploiting them. Ask for it over MCP too: “run a pen test on my site.”
Availability
The security scan is available on Pro and above, like App Test. It uses no AI, so it does not touch your Coaade AI allowance; the AI pen test is also Pro and above and uses your Coaade AI allowance, like App Test. Deploy your project, open the Security & Code review view from the IDE sidebar, and run the scan or the pen test — or add Coaade MCP to your AI tool and ask it to check your site for you.