OUR LATEST OPEN-SOURCE CONTRIBUTION

Inbox Check.
Yours to run.

A second look at suspicious email.
And the code to see how it works.

Paste a message, upload a screenshot, or connect your own Microsoft 365 mailbox. Inbox Check looks for warning signs and gives clear, consistent next steps.

MIT licensed. Use it, study it, change it, share it.

The hosted pilot is paused.

This is a software release, not an invitation to upload to RVA Cyber. Our former web checker and automated email screening at check@rvacyber.com are paused. Self-host with your own infrastructure and AI credentials; hosting and provider usage may cost money.

WHAT YOU GET

The application. Not just a demo.

01 / WEB + EMAIL

Meet the message where it is.

Text, common screenshot formats, EML, Outlook MSG and text-based PDFs in the web interface. Optional Microsoft 365 email intake uses the same assessment rules.

02 / EXPLAINABLE FLOW

Warning signs, then next steps.

Typed AI classification feeds deterministic risk rules, redacted evidence excerpts and fixed guidance. An incomplete check is labeled as incomplete.

03 / YOUR DEPLOYMENT

Your infrastructure. Your keys.

Docker packaging, setup tools, tests and operational documentation. No RVA Cyber account or agent subscription is required to self-host.

MAKE IT YOUR OWN

Start with the code.
Keep control of the rest.

You will need Linux/Docker, a compatible typed-classifier provider credential, and a supported vision endpoint for screenshots. Microsoft 365 intake is optional.

Set your own provider spending limits. Review the deployment and data-handling requirements before connecting a real mailbox.

  1. Get the source.Browse or clone the public repository ↗
  2. Bring your own setup.Follow the self-hosting guide ↗
  3. Test, adapt, contribute.Read the evaluation notes ↗

KNOW THE LIMITS

A screening tool. Not a security guarantee.

Inbox Check evaluates the content you supply. It does not authenticate the original sender, visit links, execute attachments, or replace your mail-security controls. AI assessments can be wrong. Do not submit sensitive or regulated information without evaluating your own hosting and provider arrangements.