Six commitments, published before launch
A firewall configuration is a map of everything you protect and how. These are the rules ConfigProof holds itself to, written down so you can hold us to them.
- 01
The file is deleted
Your XML is processed in memory and on an encrypted volume. It is deleted at the end of your plan retention window, or immediately when you press Delete now. Logs keep a hash of the file, never its contents.
- 02
Secrets are redacted before anything else happens
Passwords, password hashes, private keys, pre-shared keys and API credentials are stripped in the parser, before any logging, any storage and any model call.
- 03
The model never sees your configuration
It receives normalized findings that already exist, with object names wrapped in delimiters and marked as untrusted data. The provider is contracted not to train on it.
- 04
Nothing in your file is ever executed
The parser refuses external entities, entity expansion and archive bombs. The future collector will be read only and will make outbound connections only.
- 05
Isolation and an audit trail
TLS in transit, isolation per workspace, and a record of who ran what, when, with which model and which rule pack version.
- 06
The report states its own limits
Every report ends with what could not be proven without traffic logs or topology. We would rather look less certain than be quietly wrong.
Retention by plan
| Plan | Configuration and report kept for |
|---|---|
| Free | 7 days |
| Report | 30 days |
| Engineer | 90 days |
| Consultant | 365 days |
| MSP | 730 days |
| Enterprise | -1 days |
On every plan, Delete now removes the XML, the findings and any generated PDF immediately.
What the model actually receives
It receives
- Finding titles, categories and severities computed by the engine
- Object names and types, wrapped in delimiters and marked as untrusted input
- The risk factors and confidence values the engine already assigned
It never receives
- The raw configuration file or any part of it that was not turned into a finding
- Passwords, hashes, keys or any redacted value
- Your account details, your email address or your billing data
Prompt injection is treated as a real risk, because configuration comments and object names are attacker influenced text. They are delimited, labeled as data, and the model output is validated against a strict schema before it is rendered.
Subprocessors
The third parties that touch data in the course of running the service.
| Provider | Purpose | Data |
|---|---|---|
| Hosting provider | Application, database and file storage | Uploaded configurations, findings, account records |
| Stripe | Payments and subscription management | Email, billing details. Stripe holds the card, we never do. |
| Model provider | Narrative and remediation wording | Normalized findings only, under a no training agreement |
| Transactional email | Email verification and receipts | Email address |
Reporting a problem
If you find a vulnerability in ConfigProof, email us. We will confirm within two business days, we will not threaten you, and we will credit you if you want the credit.
Email the security contact