Redaction, live
Nothing sensitive ever leaves
See what NexusNest catches across every category of sensitive data. Real prompts, real redaction, before anything reaches the AI provider.
Try it yourself
Type or paste any text
See what NexusNest would redact. Runs in your browser, nothing is sent anywhere.
Help me debug this API call. Here's my setup: API Key: sk-proj-abc123def456ghi789jkl012mno345pqr678 Endpoint: https://10.0.4.55:8080/api/v1/users Auth: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.abc123 The user [email protected] (phone: +1-555-123-4567) reported the issue. Their SSN is 123-45-6789. password=xK9#mP2$vL5nQ8
Help me debug this API call. Here's my setup: API Key: [OPENAI_KEY_1] Endpoint: https://10.0.4.55:8080/api/v1/users Auth: [CREDENTIAL_1] The user [EMAIL_1] (phone: [PHONE_1]) reported the issue. Their SSN is [SSN_1]. [SECRET_1]
Detected Items
sk-proj-abc123def456ghi789jkl012mno34...Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6Ik...[email protected]+1-555-123-4567123-45-6789password=xK9#mP2$vL5nQ8This demo runs entirely in your browser using our Stage 1 regex engine. No data is sent anywhere. The full product adds Stage 2 (custom redaction rules) and Stage 3 (proprietary contextual detection).
Coverage
What we catch
Every major sensitive-data category, plus your own rules for industry-specific terms and internal codenames.
Personal Information
+1 (555) 123-4567[PHONE_1]SSN: 123-45-6789SSN: [SSN_1]
Financial Data
Card: 4532-1234-5678-9012Card: [CREDIT_CARD_1]
Account: 12345678 Routing: 021000021Account: [BANK_ACCOUNT_1] Routing: [ROUTING_NUMBER_1]
Health Data
Patient ID: MRN-2024-45892Patient ID: [HEALTH_ID_1]
HRN: H-789456HRN: [HEALTH_ID_1]
Credentials & Secrets
sk-proj-abc123def456ghi789jkl012mno345[OPENAI_KEY_1]ghp_1234567890abcdefghijklmnopqrstuvwxyz12[API_KEY_1]AKIAIOSFODNN7EXAMPLE[AWS_KEY_1]Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6...[CREDENTIAL_1]password=SuperSecret123!password=[SECRET_1]
Source Code & Keys
-----BEGIN RSA PRIVATE KEY-----
MIIEpAIBAAKCAQEA...
-----END RSA PRIVATE KEY-----[PRIVATE_KEY_1]Real prompts
Real-world scenarios
These are actual prompts employees send to AI tools every day. See what PromptWall catches, and what the AI tool actually receives.
Developer asking ChatGPT for help
Software Engineer using ChatGPT
I'm getting a connection error. Here's my config: DB_HOST=prod-db-master.acme.internal DB_USER=admin DB_PASS=xK9#mP2$vL5nQ8 DB_NAME=customers The error happens when user [email protected] tries to log in. My API key is sk-proj-abc123def456ghi789jkl012mno345pqr678.
I'm getting a connection error. Here's my config: DB_HOST=prod-db-master.acme.internal DB_USER=admin DB_PASS=xK9#mP2$vL5nQ8 DB_NAME=customers The error happens when user [email protected] tries to log in. My API key is sk-proj-abc123def456ghi789jkl012mno345pqr678.
Credit analyst uploading a loan book
Credit Analyst using Claude
From the attached loan_book_Q3.csv, list every borrower who missed two EMIs in a row. Start with Rohan Mehta, PAN ABCPM1234K, loan account 50100234567891, EMI ₹18,450.
From the attached loan_book_Q3.csv, list every borrower who missed two EMIs in a row. Start with Rohan Mehta, PAN ABCPM1234K, loan account 50100234567891, EMI ₹18,450.
DevOps engineer debugging with Cursor
DevOps Engineer using Cursor
This Terraform isn't working. Here's the relevant config:
provider "aws" {
access_key = "AKIAIOSFODNN7EXAMPLE"
secret_key = "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY"
region = "ap-south-1"
}
resource "aws_instance" "prod" {
ami = "ami-0c55b159cbfafe1f0"
instance_type = "t3.large"
private_ip = "10.0.1.42"
}This Terraform isn't working. Here's the relevant config: provider "aws" { access_key = "AKIAIOSFODNN7EXAMPLE" secret_key = "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY" region = "ap-south-1" } resource "aws_instance" "prod" { ami = "ami-0c55b159cbfafe1f0" instance_type = "t3.large" private_ip = "10.0.1.42" }
Finance team querying Gemini
Finance Analyst using Gemini
Analyze this customer refund: Customer: Sarah Johnson, [email protected] Card ending 4532-8901-2345-6789 was charged $4,500 Account routing: 021000021, Account: 1234567890 Refund reason: duplicate charge on invoice INV-2025-0892
Analyze this customer refund: Customer: Sarah Johnson, [email protected] Card ending 4532-8901-2345-6789 was charged $4,500 Account routing: 021000021, Account: 1234567890 Refund reason: duplicate charge on invoice INV-2025-0892
before it reaches the AI
Only the redacted text reaches the AI tool. The answer still works.
to protected
Single-agent install. No SASE, no policy workshop.
major AI tool covered
ChatGPT, Claude, Copilot, Gemini, Cursor, Perplexity, Grok + any HTTPS AI.
Only redacted text reaches the AI tool
Redaction is one-way. Sensitive values are replaced with placeholders in flight, before the request leaves for the AI tool.
No token maps or reverse-mapping capability exists. Even we cannot reconstruct the original data.
Audit logs store the redacted version only: what was detected, which categories, when. Not the original content.
Non-AI network logs contain metadata only: URL, domain, method, status, timing. No request or response bodies.
DPDP Act aligned by design. The AI tool only ever receives the redacted version of a prompt or file, and only the redacted version is persisted.