Secure the apps you ship. Without slowing down.
Find the apps you own, read them from source, test them in a real browser fully authenticated, and close the loop with fixes your team approves.

Apps multiplied. Testing did not.
Traditional DAST leaves five gaps.
New to this? Read our guide to application security testing.
Capabilities that set us apart from traditional DAST.
Before the scan
- App discovery finds the apps tied to your domain, including ones nobody onboarded.
- Source intelligence maps the routes your frameworks declare: API discovery from source code.
- Your pipeline or your coding agent starts the scan.
The scan
- Routes a crawler would never click still get tested.
- A real-browser DAST crawler scans fully authenticated, TOTP-based MFA included.
- AI endpoints get their own tests.
After the scan
- Your coding agent drafts the fix. Your team approves it.
- A rescan proves it is fixed.
- False positives you mark stay suppressed in future scans of that app.
One loop, starting and ending in your source code.
App discovery feeds the loop. Source intelligence plans the API scan, a rescan proves each fix, and your triage carries forward.

Source intelligence studies the app
Source intelligence reads your code before the scan and maps the API routes your frameworks declare, the parameters they take and, for supported frameworks, how each route is protected. Deterministic and local; no LLM reads your code.
API discovery from source code →
The scan runs the plan
One engine for web apps and APIs. A real-browser crawler fills forms, waits for JavaScript and scans signed in; API scans test the routes from the source map, including the ones a crawler would never click. AI endpoints get their own tests.
Web app DAST → AI endpoint security testing →Remediation and learning
Fix it: your coding agent drafts the change and your team approves it. Prove it: a rescan shows what is fixed, new and still open. Carry it forward: false positives you mark stay suppressed in future scans of that app.
Remediation and learning →That was one application.
The same loop works for each app you turn into a target, including the ones app discovery finds.
Find the apps nobody onboarded.
Start from your domain to find the web apps and APIs tied to it, including ones nobody onboarded. Confirm which ones are yours, then turn them into scan targets.
Start from your domain
Certificate transparency logs, DNS and archived web history point to the hosts tied to your domain.
See why each one is listed
Each app shows which source found it and how it relates to your domain, so you confirm ownership before anything is tested.
Turn them into targets
Pick the apps you confirm and create scan targets from the inventory.
Available on Enterprise. We turn it on for your proof of concept.
AI endpoints get their own tests.
NightVision tests whether an endpoint backed by a language model can be steered into calling out to an attacker or reading your cloud metadata service, and reports those only with proof.
Calls out to an attacker
The model is induced to fetch a one-time callback address. A finding needs that callback to arrive.
Reads your cloud metadata
The model is prompted toward the cloud metadata service. A finding needs proof the response came from it.
How the cloud metadata check works →Flags model-backed endpoints
NightVision also flags endpoints it recognizes as model-backed, so you can see where AI features sit in the app.
Every finding brings evidence. Every fix gets rescanned.
Reproduce it
The exchange that proved it, plus a one-command curl repro.
Trace it
File and line on supported API findings, not a category.
Verify it
Green means it no longer reproduces.
In a head-to-head evaluation at a Fortune 500 healthcare company, NightVision scans finished 51% faster than the incumbent enterprise scanner, with fewer false positives.
Agents are only as good as the scanner they call.
Start a scan from your pipeline on every pull request or deploy, or from the coding agent you already use. NightVision supplies the evidence, and a rescan proves the fix before a human merges.
Works with any MCP-capable coding agentClaude Code · Codex · Cursor · GitHub Copilot
Determinism on the finding, agents on the fix, humans on the merge.
"Our team won an internal hackathon award using NightVision."Steve McKinnon · BeyondTrust
Questions buyers actually ask.
What makes NightVision's DAST crawler different?
It drives a real browser with three mechanisms built for modern apps: LLM-based form handling that fills forms with valid, internally consistent values; WebDriver BiDi-based intelligent waiting that advances only when network and DOM activity quiesce; and duplicate page detection that collapses templated pages into one logical state.
Can NightVision run fully authenticated scans with MFA?
Yes. Record your login once with Playwright and NightVision replays it at scan time, with credentials automatically vaulted and TOTP/MFA supported. A login check verifies authentication before testing begins and fails the CI pipeline instead of silently scanning logged out.
Can NightVision scan undocumented APIs?
Yes. For supported REST frameworks, source intelligence (API eNVy™) reads how your code declares its routes and writes an OpenAPI spec in under 20 seconds. It runs locally, no LLM reads your code, and recognized shadow and undocumented routes feed your API scan alongside your web app scan.
Can NightVision test AI and LLM endpoints?
Yes. NightVision tests whether an endpoint backed by a language model can be steered into calling out to an attacker or reading your cloud metadata service, and reports those only with proof. It also flags endpoints it recognizes as model-backed. These tests are part of NightVision's standard DAST check set.
Does NightVision find apps we have not onboarded?
Yes. App discovery starts from your domain to find the web apps and APIs tied to it, including ones nobody onboarded. You confirm which ones are yours, then turn them into scan targets. App discovery is available on Enterprise, and we turn it on for your proof of concept.
Will a false positive I dismiss come back on the next scan?
No. Mark a finding false positive or won't fix once and NightVision carries that triage forward, so the same finding on the same endpoint does not come back as a new open finding in future scans of that app.
Can coding agents use NightVision?
Yes. Through NightVision's public MCP server and open-source Agent Skills, coding agents can launch scans, inspect runtime evidence, use source context where available, propose changes, and rescan. Repository controls and human review still govern the merge.
How does NightVision handle vibe-coded and AI-generated applications?
AI-assisted development creates applications faster than security teams can inventory them, often outside traditional pipelines. NightVision scans any running web app or API, discovers endpoints from source where a repository exists, and lets coding agents launch scans as part of the build itself, so new applications get tested instead of slowing down.
Is there a free trial?
Yes. NightVision offers a self-serve free trial with no credit card required. Paid plans start at $15,000 per year for a single application with unlimited seats, and custom enterprise pricing is available for teams.
Recent writing on application and API security
Your AI Endpoint Can Reach Your Cloud Credentials: proving it, not guessing it
An AI endpoint can be steered into reading the cloud metadata service. NightVision confirms the read by the provider's own signature, not the model's reply.
InsightCSP vs. Subresource Integrity, Explained: What Each Stops When a Vendor Script Is Tampered
CSP limits where scripts load from; SRI checks the exact bytes. How each held up when Brevo's embedded scripts were altered at the CDN edge.
InsightOpenAPI Extensions for Security Testing: x-source and x-nv-dast-sinks Explained
Two OpenAPI extensions for security testing: x-source records where a route is declared; x-nv-dast-sinks lists the SQL, shell, and HTTP calls it can reach.
InsightThe Callback That Arrives After the Scan: What Out-of-Band Testing Forgets
An out-of-band check is proven by a callback, and the target decides when it arrives. A scanner that stops listening when its check ends misses the proof.
InsightSources and Sinks, Explained: The Model Behind Every Injection Vulnerability
Sources and sinks explain SQL injection, XSS, SSRF, and command injection with one model. Why the sanitizer must match the sink, and what counts as proof.
InsightDAST False Positives That Stay Suppressed: Stable Finding Identity Across Runs, SARIF, GitLab and Jira
Why DAST false positives come back every scan, and how a stable finding identity keeps them suppressed and dedups SARIF, GitLab and Jira across runs.
See what your applications actually expose.
Map it from source and the browser, test it fully authenticated, and get evidence your team and agents can act on. Self-serve trial, no credit card. SOC 2 Type 2.
Next step: a proof of concept on your apps
Bring the apps you know. We turn on app discovery to find the ones you do not.
Web apps and APIs, with source intelligence where you share a repository.
Findings with evidence, fixes your agent drafts, and a rescan that proves them.
Evaluating DAST tools? See how NightVision compares to Checkmarx, Bright Security, and other DAST and API security scanners.