MCP privacy: mask people in AI responses
When you connect an AI assistant to In Parallel over MCP, it can read your meetings, goals, and Findings. MCP privacy lets you share that context without sharing who's who: turn it on and In Parallel replaces identifiable people with stable pseudonyms before any response reaches your assistant.
The assistant can still tell people apart and follow the work — it just can't see real names or email addresses.
What it does
The Mask people in MCP responses control rewrites every MCP response on its way out of In Parallel. Each real person becomes a consistent pseudonym — something like Person 4F2A1C — and their email becomes a masked token like [email protected].
Because the same person always maps to the same pseudonym — across every tool and every response — your assistant can still reason about the work ("Person 4F2A1C owns three open action items") without ever learning the real identity behind it.
What gets masked
Masked | Examples |
People's names | Full names and the first names of people In Parallel knows — including where they're mentioned in summaries and notes |
Email addresses | Replaced with a masked token tied to that person's pseudonym |
Meeting speakers | Speaker labels on transcripts and meeting records |
Phone numbers | Masked in the fields that carry them |
What isn't masked
Masking is deliberately narrow so your data stays usable. The following are not changed:
The In Parallel product and the In Parallel recorder — these are not people, so the company name and the meeting-recorder bot are left intact rather than turned into pseudonyms.
Workspace, team, and object names — a Workspace called "Back-to-School Campaign" stays readable; only people are masked.
IDs, URLs, and filenames — identifiers are left untouched so they keep working, even on the rare occasion a filename contains a name.
People outside your In Parallel directory — In Parallel only masks people it already knows: members of your workspace and people who've taken part in your recorded meetings. A name that appears in free text but belongs to no one In Parallel knows may not be caught (it doesn't guess at arbitrary names).
Turn it on
MCP privacy is a per-user setting — it applies to your connected AI tools, and each person controls their own.
Open Personal settings and go to the Integrations tab (the same place you manage MCP connectors).
Find the MCP privacy section.
Switch Mask people in MCP responses on.
From then on, every MCP response to your assistant is masked until you switch it back off.
Rolling out: MCP privacy is an early-access control and may not be enabled for every tenant yet. If you don't see the MCP privacy section in your settings, contact your In Parallel customer success representative to activate it.
Another way to control what you share
MCP privacy governs what your connected AI tools see. If instead you want to take a single meeting's full record somewhere yourself, you can download that meeting's raw transcript from the meeting page — a non-MCP way to choose exactly what data leaves In Parallel and where it goes.
Related
Connect your AI tool to In Parallel — set up Claude, ChatGPT, or Cursor over MCP.
In Parallel MCP tools reference — the tools whose responses masking applies to.
Download a meeting's raw transcript — the non-MCP way to control your data.
Governance, security, and trust — how In Parallel handles access and data.
AI in In Parallel — what the AI layer does, and what it doesn't.