AI & privacy

MCP PII redaction: AI agents without personal data

MCP PII redaction applies where agents fetch data: emails, tickets, databases. Before a tool result reaches the model, Diskretio replaces names and ID numbers with placeholders, via REST API today, with an MCP connection in preparation.

anonymize.shresponse.json

# Anonymize a file, report as JSON

$ curl -X POST "$API/v1/files?response=json" \

-H "x-api-key: $ANON_KEY" \

-F "file=@Employment_Contract_Smith.pdf" \

-F 'options={"strategy":"placeholder","language":"en"}'

HTTP/1.1 200 OK · x-pages-charged: 3 · x-pages-remaining: 1757

{

"filename": "anonymized.pdf",

"pages": 3,

"report": {

"counts": {

"PERSON": 6,

"ORG": 2,

"ADDRESS": 3,

"IBAN": 1,

"TAX_ID": 1,

"NATIONAL_ID": 1,

"EMAIL": 1

},

"entities": [{ "id": "PERSON_1", "label": "PERSON", "occurrences": 4 }, …],

"warnings": []

},

"content_base64": "JVBERi0xLjcKJeLjz9MKMSAwIG9iago8PC9UeXBl…"

}

# page 1 of the output

Between [ORG_1] and [PERSON_1], [ADDRESS_1] …

Your formats and the AI tools you work with afterwards

  • Word
  • Excel
  • PowerPoint
  • Outlook (emails)
  • PDF
  • Scans & photos
  • TXT Text & CSV
  • ChatGPT
  • Claude
  • Microsoft Copilot
  • Gemini
  • Mistral
  • DeepL
  • Perplexity

Where agents collect personal data

  • Tool results from CRM, ticketing or databases
  • Files read from drives and document stores
  • Email and chat content an agent summarizes
  • Web pages and forms with contact details
  • Logs and intermediate results the agent stores
  • Content passed on to another tool or agent

How it works

  1. 1

    Plan an anonymization step

    Define which tool results may contain personal data and route them through Diskretio before they reach the model.

  2. 2

    Call the API

    Send text to the text endpoint and files to the file endpoint. The response contains cleaned content and a report of the categories found.

  3. 3

    Agent works with placeholders

    The agent plans and responds based on [PERSON_1] and [EMAIL_1]; real values stay in your systems.

Why agents are a privacy problem of their own

In a classic chat, a human decides what to type. An agent, by contrast, uses the Model Context Protocol (MCP), an open standard for connecting tools and data sources to language models, to call tools on its own and receives their results unfiltered in its context. A ticket export or mailbox thus lands in full at the model provider.

The cleanest fix is a fixed anonymization step between data source and model. It works regardless of which model the agent uses.

Today: REST API, MCP connection in preparation

Diskretio offers a REST API with endpoints for text and files. You can build it into your own MCP servers, agent frameworks or tool wrappers: the tool fetches the data, calls our API and only returns the cleaned result to the model.

A dedicated Diskretio MCP server that agents can add directly as a tool is in preparation. We will announce it once it is available.

Limit: no automatic re-identification

We store no mapping between placeholder and original value. An agent can analyze, classify and draft with anonymized data, but cannot itself send an email to the real address behind [EMAIL_1]. For such actions, your system must insert the original value from its own source.

FAQ

Questions. Answered briefly.

Try it with your own document

10 pages per month are free. No subscription, no credit card.