01
Choose the scope
We agree on the project, security checks and the materials needed for a pilot.
Security checks.
Verified fixes.
GuardFix checks dependencies, secrets, code and configuration.
The beta agent prepares npm updates for verification and engineering review.
Beta product. Guided demos and scoped pilots. Check scope and limitations
GuardFix workspace
GuardFix
AI analysis Findings Project Tasks and fixesMy account
Findings and next steps
YouWhat should we fix first?
GuardFix
The scan found a dependency with a known vulnerability. Check whether the affected feature is used, review the fixed version and test the update before release.
Needs reviewA dependency needs a security update
Check exposureReview the fixTest the update
FOR BUSINESSES
THAT RELY ON SOFTWARE
THAT RELY ON SOFTWARE
SaaS teams
AI applications
Online businesses
Small product teams
One project. Several layers of checks.
Dependencies, secrets, code and configuration.
A clearer starting point for your security review.
Dependency checks
Find known vulnerabilities in supported project dependencies and identify versions to review.
Secret detection
Look for possible credentials in project files and get guidance on investigating exposure.
AI-assisted triage
Understand scanner findings, review priorities and turn evidence into tasks for your team.
Code checks
Spot selected unsafe code patterns with Opengrep and GuardFix rules.
Configuration review
Review supported configuration files for security-relevant settings and misconfigurations.
Actionable reports
Bring findings, evidence and suggested remediation together for an engineering review.
From a security finding
to a checked fix.
YouWhat should we fix first?
GuardFix Agent
The scan found a dependency with a known vulnerability. Check whether the affected feature is used, review the fixed version and test the update before release.
Check exposureReview the fixTest the update
Review the affected dependencyReady for engineering review
AI prepares the update.
Verification decides if it passes.
GuardFix brings findings, priorities and remediation tasks into one system. The beta agent prepares updates for vulnerable npm dependencies. A separate verifier checks installation, target vulnerabilities, builds and tests.
A fix is ready only when the required checks pass. An engineer reviews the evidence and agrees changes with you before they are applied to your project.
TrivyDependencies & configuration
Gitleaks · OpengrepSecrets & code patterns
Separate verifierInstall, rescan, build and tests
The agent proposes a fix.
A separate verifier checks it.
Our first measurement covers 30 npm upgrade tasks from Snyk VulnBump. GuardFix passed all required checks on 22 tasks: installation, a rescan of the target vulnerabilities, build and tests.
These are early internal results on two open-source projects. We are preparing an evaluation on tasks the agent was not developed against.
About Snyk VulnBumpResults on 30 npm tasks
Snyk VulnBump · two repositories
| System and model | Tasks | Pass |
|---|---|---|
| GuardFixSonnet 5 | 22/30 | 73% |
| Snyk EvoSonnet 5.5 | 21/30 | 70% |
| Claude Code unaidedSonnet 5.5 | 14/30 | 47% |
| Package manager bumpNo AI | 9/30 | 30% |
Preliminary results from one attempt per task. GuardFix used Sonnet 5 and was developed on these same 30 tasks; the published Evo and Claude Code runs used Sonnet 5.5. Different conditions and a one-task gap to Evo do not establish superiority. Held-out and full-benchmark results are pending.
A report you
can act on.
See what was detected, where it was found and what to investigate next. Keep the findings tied to evidence rather than a score alone.
- Scope, scanner coverage and limitations
- Findings with evidence and severity
- Remediation tasks and verification results
Illustrative example · no customer data
GUARDFIXProject security review01 / 04
Risk overview
Illustrative report| Priority | Finding | Next step |
|---|---|---|
| High | Secrets | Review and rotate the key |
| Medium | Dependency findings | Review the fixed version |
| Low | Code | Validate input handling |
- SecretsHigh
Next stepReview and rotate the key
- Dependency findingsMedium
Next stepReview the fixed version
- CodeLow
Next stepValidate input handling
Next steps for your team
From findings to checked fixes.
A guided pilot, with engineering support.
Scope and access are agreed before work starts.
02
Run the checks
Our scanners check project files for vulnerable dependencies, exposed secrets, code patterns and configuration issues.
03
Prepare the update
AI explains the findings and prepares npm dependency updates. The original findings remain available for review.
04
Verify and review
A separate verifier checks the update. An engineer reviews the evidence and agrees the changes with you.
Start with a focused pilot.
See the system in a demo, then agree on a project and a measurable outcome. We define scope, access, human review and pricing before paid work begins.
Before you
get started.
Scope, access and what to expect.
What is available in the beta?
Guided demos and scoped pilots. Our team runs the checks, reviews findings and discusses the result with you. The current workflow checks project files for dependency, secret, code and configuration issues; the remediation agent currently prepares npm dependency updates.
Do we need an in-house security team?
No. We explain the findings and help define the next steps. Engineering support and the scope of remediation are agreed for each pilot. Requesting a demo does not require production access or installing software on your servers.
Does the AI find every vulnerability?
No tool can guarantee that. Scanners produce findings; AI helps explain and prioritise them. Results can include false positives and miss issues, particularly runtime and business-logic vulnerabilities. Review the evidence before making changes.
Which tools do you use?
The current analysis pipeline combines Trivy, Gitleaks and Opengrep with GuardFix rules and an AI explanation layer. Each tool has a different scope. A failed or incomplete check must not be treated as a clean result.
How are fixes checked and applied?
The agent prepares an npm update. A separate verifier checks a clean install, target vulnerabilities, builds and tests. Failed or unavailable checks are reported. An engineer reviews the evidence and agrees changes with you; production changes require your approval.
Can we start with a public repository?
Yes. A public open-source project or a dedicated sample is a practical starting point. Avoid uploading real secrets or personal data. Any use of private customer projects needs an agreed scope and access arrangements.
What happens after I request a demo?
We review your request, contact you by email and arrange a walkthrough. Application access is provided separately by invitation. Submitting the form does not create an account or start a paid subscription.
How do pilots and pricing work?
After the demo we agree on the project, expected outcome, materials, engineering support and price. There is no automatic charge or subscription when you request a demo.
Is this a replacement for a penetration test?
No. Project scanning helps identify particular classes of issues. A penetration test, runtime review or specialist assessment may still be needed for your application and threat model.
See GuardFix
in action.
Book a guided demo. We will show the workflow and discuss your project.