Skip to main content
Masker is a compliance firewall for voice AI. It proxies every conversation turn between your voice platform and your LLM, detects and tokenizes protected health information before it reaches the model, and reverses the substitution on the way back — so your callers hear natural responses while your LLM provider’s logs never contain real PHI.
Status: private beta. The portal at masker-voice.fly.dev and the live demo at try.masker.dev are open for evaluation today. Production beta — with a signed BAA, SLAs, and VPC deployment — launches May 30, 2026. To join, see Beta access.

​
What you can do with Masker

Connect a voice agent

Sign in with GitHub, create an agent in the portal, and get a per-agent proxy URL in under two minutes. No SDK, no code change — just a URL.

See live redaction

Place a real call and watch PHI get tokenized before it reaches OpenAI. Verify in OpenAI’s own completion logs that no raw data crossed the boundary.

Download signed audit reports

Every session generates an Ed25519-signed compliance report with a full BAA chain, HIPAA Safe Harbor coverage breakdown, and a verifiable audit chain.

Try the no-signup demo

The hosted demo at try.masker.dev runs the full pipeline against a stub LLM — no account or Vapi key required.

​
Feature status

These are the capabilities available today, in beta, or on the roadmap. Every feature listed here reflects what the product actually does — not a marketing projection.

​
Where to start

If you’re new to Masker, start with the Quickstart to place your first masked call. If you want to understand the proxy pipeline before touching configuration, read How it works first. For production access or a pilot, see Beta access or email hello@masker.dev.