AAVARAN · DATA MASKING
Mask sensitive data before any agent or model sees it.
AgentAnywhere Aavaran (आवरण, a covering) is the data-masking and de-identification control for AI agents. It classifies PII, financial, and regulated data on each call it handles and masks it — reversible tokenization or irreversible redaction — before a model ever sees the raw values. Deploy it standalone or natively in the platform, with every mask and unmask logged as a Trust Receipt.
Hi, I'm Anita Rao. My email is a***@e***.comemail and my card ************1111card was charged twice. My PAN is AB***FPAN.
Watch it happen: only the masked form crosses the line.
The hero above is not a diagram — it is Aavaran's own detection and masking engine running on a support message. Here is the same idea, larger: sensitive values are found inline and replaced with masks that keep only a small recognisable part — the last four digits of a card or phone number, the first letter of an email's name and of its domain, the first two and the last character of a PAN. Other fields are masked in full. Only the masked form crosses your perimeter to the model. Aavaran is designed so that raw data does not leave your boundary.
Hi, I'm Anita Rao. My email is a***@e***.comemail and my card ************1111card was charged twice. My PAN is AB***FPAN.
Raw data should never reach the model.
Agents touch sensitive data on nearly every call — customer names and contacts, account and card numbers, health records, anything your regulators consider protected. The moment that data leaves your boundary for a model, an embedding service, or a tool, you have created an exposure you cannot take back.
Prompt redaction bolted on per application does not scale, and it cannot be audited. There is no single place to prove what was masked, what was sent in the clear, or who reversed a mask and why.
Aavaran makes masking a control, not a coding convention: one place that classifies sensitive fields, masks them before egress, and records the evidence.
Classify. Mask. Enforce. Prove.
Aavaran is designed to sit in front of the model, as an OpenAI-compatible proxy or as a mask-before-inference call that a gateway or agent runtime makes. It detects sensitive data, masks it according to policy before anything reaches the model, and writes a signed record of what happened.
Classify
Detect PII, financial, and regulated fields inline, on each request it handles — names, contacts, account and card numbers, health identifiers, and the fields your policies designate.
Mask
Apply reversible tokenization or irreversible redaction per field. Tokens are deterministic — the same value always gets the same token — so downstream systems can still match records without ever seeing raw values. A format-preserving token scheme is available as an option, and reversible tokenization is switched on per deployment, not by default.
Enforce
Aavaran is designed to sit in front of the model, as an OpenAI-compatible proxy or as a mask-before-inference call that a gateway or agent runtime makes, so that raw data does not reach a model unmasked, with every mask and unmask logged. It applies the policy created in Aavaran.
Prove
Every mask and unmask is logged as a Trust Receipt — who, what, under which policy, when — signed and versioned, exportable for auditors.
Reversible when you need it back. Irreversible when you don't.
Masking is per field, by policy. Choose the mode that matches the data and the obligation — and prove which one was used.
Tokenize — reversible
- Deterministic tokens stand in for raw values; a format-preserving scheme is available as an option
- Downstream joins, lookups, and analytics keep working on the tokens
- Unmasking is policy-gated and itself logged as a Trust Receipt
- For data you must be able to restore under a controlled, audited path; switched on per deployment, not by default
Redact — irreversible
- In the full style nothing of the value remains; the default partial style keeps the small part described above, and the rest cannot be recovered
- In the full style, nothing to leak, subpoena, or reverse later
- For data the model never needs and you never want to retain
- The default for the most sensitive fields under the strictest policies
Buy it alone. Or let the platform run it.
Aavaran is the lightest first purchase and a hard RFP checkbox — it stands on its own. It also disappears into the platform, where the other modules give it policy, enforcement, and a deployment posture.
Standalone
- Deploy Aavaran in front of your existing agents and models
- Define masking rules and enforce them at your egress point
- Trust Receipts give you the audit trail from day one
- No platform migration required to satisfy the masking requirement
Platform-native
- The masking policy is created in Aavaran
- Aavaran sits in front of the model: an OpenAI-compatible proxy, or a mask-before-inference call a gateway or agent runtime makes
- Swaraj is designed to run the whole thing inside your boundary, air-gapped
- Observe keeps the log; Aavaran signs its own Trust Receipts
Built for the controls your auditors ask about.
Data minimization and de-identification are expectations in every serious AI governance review. Aavaran gives you a defensible answer: sensitive data is masked before it leaves your boundary, and each mask is recorded in a signed Trust Receipt.
SOC 2 and ISO 27001 are held certifications. HIPAA and GDPR readiness is independently assessed. Aavaran is designed to support India's DPDP Act and the RBI FREE-AI framework. These describe the AgentAnywhere platform and how ShepHertz operates it; they are not a certification of your own deployment.
Further reading.
Three longer pieces, each written from a different chair. For government departments: Citizen data and AI in government: mask it before the model reads it. For banks, lenders and insurers: Customer data and AI in Indian banking: a promise is not a control. For contact centres and outsourcers: AI in the contact centre: keep card and ID numbers out of the model.
FAQ
Frequently asked questions.
- What is AgentAnywhere Aavaran?
- AgentAnywhere Aavaran is a data-masking and de-identification control for AI agents. It classifies PII, financial, and regulated data on each call it handles and masks it — by reversible tokenization or irreversible redaction — before any model, embedding service, or tool sees the raw values. Aavaran deploys standalone or natively as part of the AgentAnywhere platform.
- How does Aavaran mask data?
- Aavaran classifies sensitive fields inline, then applies masking per field according to policy. Reversible tokenization, which is switched on per deployment, replaces a value with a deterministic token that downstream systems can keep using to match records, with a format-preserving token scheme available as an option; irreversible redaction replaces the value so that it cannot be recovered: in the full style nothing of it remains, and the default partial style keeps only the small part described on this page. Aavaran is designed to sit in front of the model, as an OpenAI-compatible proxy or as a mask-before-inference call that a gateway or agent runtime makes, so that raw data does not reach a model unmasked, with every mask and unmask logged.
- What is a Trust Receipt?
- A Trust Receipt is Aavaran's audit artifact. Every mask and unmask is logged — who, what, under which policy, and when — then signed and versioned. Trust Receipts can be exported as JSON and PDF for auditors.
- Can Aavaran be deployed standalone?
- Yes. Aavaran is designed to stand on its own in front of your existing agents and models, with its own masking rules and Trust Receipt audit trail — no platform migration required. It is also designed to work inside the platform, in front of the platform's models, as an OpenAI-compatible proxy or as a mask-before-inference call that a gateway or agent runtime makes; Swaraj is designed to run it air-gapped, and Observe keeps the log.
- Which certifications and frameworks apply to Aavaran?
- SOC 2 and ISO 27001 are held certifications. HIPAA and GDPR readiness is independently assessed. Aavaran is designed to support India's DPDP Act and the RBI FREE-AI framework. These describe the AgentAnywhere platform and how ShepHertz operates it; they are not a certification of your own deployment.
Mask the data before it ever leaves your boundary.
Aavaran is the lightest way to close the data-exposure gap in your agent stack — standalone today, platform-native when you want Custodian policy and Gateway enforcement behind it. Tell us what your agents touch.