Masker works with ElevenLabs Conversational AI by replacing your agent’s LLM endpoint with Masker’s OpenAI-compatible proxy. ElevenLabs sends each user turn to Masker, which tokenizes PHI, calls your upstream model with the masked prompt, and rehydrates tokens in the response before returning it. From the caller’s perspective nothing changes — from your LLM provider’s perspective, real patient data was never there.Documentation Index
Fetch the complete documentation index at: https://docs.masker.dev/llms.txt
Use this file to discover all available pages before exploring further.
The steps below use the public demo at try.masker.dev. For a production deployment with your own Masker portal account and a dedicated agent ID, follow the Quickstart to create an agent first, then substitute that agent’s URL wherever the demo URL appears below.
Set up the integration
Get your session URL
Open try.masker.dev. The dashboard generates a session token for your browser tab. Your Masker URL for ElevenLabs looks like this:Copy the full URL — the
/s/<your-session>/ segment pins all requests from this tab to the same session in the Masker dashboard.Open your ElevenLabs agent settings
Go to the ElevenLabs Conversational AI dashboard and open the agent you want to route through Masker.
Switch to Custom LLM
In the agent settings, find the LLM section. Change the LLM provider to Custom LLM. A Server URL field appears.
Paste your Masker URL
Paste the URL you copied in step 1 into the Server URL field:ElevenLabs appends
/chat/completions to this base URL automatically — do not add it yourself.Leave the API key blank
The public demo does not require an API key. Leave the API key field empty. Authentication is handled by the session token embedded in the URL path.
In a production deployment using your own Masker agent, authentication still happens via the URL path — you do not need to set an API key in ElevenLabs for any Masker configuration.
Verify redaction is working
While a conversation is running, switch to the Masker dashboard tab. The transcript view shows two sides of the firewall:- Left (regulated): the raw transcript your caller spoke, including any PHI.
- Right (public): the masked version sent to your LLM, with PHI replaced by tokens like
[SSN_01]or[PHONE_02].
/s/<your-session>/ segment.
Demo URL vs. production URL
| Environment | URL format |
|---|---|
| Public demo | https://try.masker.dev/agents/elevenlabs-public-demo/s/<your-session>/v1 |
| Production (portal) | URL shown in the Masker portal under your agent’s settings |