Product guide · By Aurelio Avila · October 30, 2026
Remove personal data and secrets from a prompt before AI sees it
You want an AI chat to rewrite an email, summarise a ticket or debug a log. It needs the problem, not your client's phone number or a live API key. Redaxa checks the text first, marks what looks sensitive and gives you a redacted copy. You read it and decide what to send.
Install the free Chrome extensionTry a free prompt check
The extension's free checks run in your browser: the text stays on your device. The web check runs on Redaxa's servers, with 5 free checks a day and no account.
Before and after: a made-up example
Every name, number and key below is invented. The redacted version is what Redaxa's engine returns for this text.
Before
Hi Laura, can you summarise this for the client? Call me on +44 20 7946 0958 or email laura.bianchi@example.com. The test key is sk-proj-RedaxaDocsExampleNotARealKey1234567890 and the refund goes to IBAN GB82 WEST 1234 5698 7654 32.
After
Hi [NAME], can you summarise this for the client? Call me on [PHONE] or email [EMAIL]. The test key is [SECRET] and the refund goes to IBAN [IBAN].
Five findings: a personal name, a phone number, an email address, an API key and an IBAN. The request itself is untouched, so the AI can still do the job.
What Redaxa detects
The Chrome extension, the web check and the Windows app share one engine. Each type below has its own page with real engine output, near misses and known false positives.
- Personal data: email addresses, phone numbers, names after a greeting, English-format street addresses, IPv4 addresses, US Social Security numbers, UK National Insurance numbers and Italian fiscal codes.
- Financial data: card numbers (Luhn-checked), IBANs (mod-97-checked) and crypto wallet addresses.
- Credentials: API keys and tokens such as OpenAI, Anthropic, GitHub, Stripe and AWS, plus bearer tokens, JWTs, passwords, private keys and secrets in configuration.
- Your own words: protected terms such as client and project names, which no generic pattern can know.
The full list is in the detection reference.
Where each check runs
| Where you check | Where the text goes | Free use |
|---|---|---|
| Chrome extension, free checks | Nowhere. The check runs in your browser on ChatGPT, Claude, Gemini and Perplexity, before the message is sent. | No account, no daily limit. |
| Web check and Windows app | Redaxa's servers, over an encrypted connection. The text is analyzed and discarded; it is not stored and not forwarded to an AI provider. | 5 checks a day, no account. |
| Extension with Pro or Business | Redaxa's servers, the same way, so your protected terms and team policies apply. | Paid plans, no daily limit. |
Activity records, where enabled, hold metadata only: time, application, finding categories and counts. Never the text. See the privacy policy.
Limits
- Detection can miss sensitive information or flag harmless text. A clean result means the checks found nothing, not that the text is safe.
- A name is detected only after a greeting such as "Hi" or "Dear". Add the names that matter to you as protected terms.
- Street addresses are recognised only in the English house-number-first format.
- The phone rule also matches some order and invoice numbers.
- Redaxa finds possible credentials; it never tests whether a key works. If a real key was already shared, rotate it: redacting a copy does not revoke it.
Start free
- Add Redaxa from the Chrome Web Store, open your AI chat and press Enter as usual: Redaxa checks the message before it leaves.
- Or paste text into the free prompt check on the web and copy the redacted version.
Install the free Chrome extensionTry a free prompt check
Pro (€7.99 a month or €79.90 a year) removes the daily limit and adds protected terms in the extension. Pro and Business start with a 3-day free trial: a card or another payment method is needed, nothing is charged for 3 days, and you can cancel before it ends from Manage subscription. Prices exclude VAT. See plans.
FAQ
Does Redaxa send my prompt to an AI provider?
No. Free checks in the Chrome extension run in your browser and the text stays on your device. Checks on the web, in the Windows app and with a Pro or Business plan run on Redaxa's servers: the text is sent over an encrypted connection, analyzed and discarded. It is not stored and not forwarded to an AI provider.
Does it change my prompt without asking?
No. You see what was found and a redacted copy. In the extension you choose to replace the message with the redacted version, send it anyway, or close the notice and edit it. On the web you copy the redacted version yourself.
Is it free?
Yes, for the basic checks. The Chrome extension's free checks have no daily limit and need no account. The web check and the Windows app give 5 free prompt checks a day without an account. Pro and Business remove the daily limit and add protected terms in the extension and team policies.
Will it find every piece of sensitive data?
No tool can promise that. Detection can miss sensitive information or flag harmless text, so read the redacted version before you send it.
Which AI chats does the extension work on?
ChatGPT, Claude, Gemini and Perplexity. For any other tool, check the text on the web or in the Windows app and paste the redacted copy.