How is this better than pseudonymizing and pasting?

It is a question of order. Copy-paste works, as long as you do it every time and before you have shown the file. The connector removes the condition: the model asks for the document and already receives the anonymized version. There is no moment at which the original text sits in the conversation.

Will my AI actually use it?

Not on its own every time, and that is a connector's real weak point. Say it once in your standing instructions: when I give you a document that may hold personal data, read it with lire_document_anonymise rather than with your usual file tool. Once a file has been read normally, the names are in the conversation and nothing takes them back out.

Connect your AI

Let Anoni read the document first.

Your AI requests a file. Anoni reads it on your machine and returns text with placeholders in place of detected personal data.

Set up the connector Step-by-step guide

Lease excerpt · fictional example

The lease stays on your machine.

The landlord, Ms Camille Fontaine, residing at 14 rue des Lilas in Nantes, lets the property described below.

Fictional example. Give the connector the file path without attaching the original to the conversation.

Capabilities

What your AI can request.

The result depends on the action. A copy created on your device is not automatically sent to your AI.

Find personal data

Says what personal data a document holds, and how much of each kind. Transforms nothing, writes nothing.

AI receivesCounts per category. Not one line of the document.

Read an anonymized version

Returns the document's text with every piece of personal data replaced by a marker. The reading tool, instead of its own.

AI receivesThe anonymized text, and nothing else.

Create a pseudonymized copy

Writes a copy that keeps the layout. Word and Excel come back in their own format, a PDF keeps its pages.

AI receivesThe path of the file that was written.

Redact a file

Writes a copy with real black bars. The hidden text is removed from the file, not merely covered.

AI receivesThe path of the file that was written.

Installation

Set up once. Then bring your documents.

The connector is included in the app downloaded from this site. Works with Claude Desktop, Claude Code, and MCP clients.

Copy the configuration from Anoni
Settings, General tab, “Connect an AI”, then “Copy the configuration”. The connector is already in the app: there is no file to go and find.
Paste it into your client
Claude Desktop, Claude Code, or any other MCP-compatible client. Open its configuration file, paste, save, restart the client.
Leave Anoni open
The connector goes through Anoni's engine, on your machine. With the app closed it says so and stops, instead of reading the file itself.

{
  "mcpServers": {
    "anoni": {
      "command": "node",
      "args": ["/path/resolved/by/Anoni/anoni-mcp.mjs"]
    }
  }
}

Example only: copy the configuration from Anoni for the correct path on your machine. With AppImage, the path changes at each launch.

Worth knowing

Deliberately limited access.

No AI decoding or mapping table. The connector and the app have different capabilities.

The encrypted vault that restores your data is sealed by the app, not by the connector. Your AI cannot decode anything, and will not offer to.

Signatures, stamps, and photos are not text and are not included in counts. Anoni flags them to the AI and covers them during redaction.

Questions

Before connecting.

A standard that lets an AI call tools installed on your own machine. Your AI client starts Anoni's connector when it launches and talks to it locally. The document is processed on your machine. The tool result is then returned to your AI client; if it uses a hosted model, that result may be sent to its provider.

Not through this route. The connector writes no vault, so nothing it produces can be decoded. For the full round trip, open the document in Anoni: the mapping stays sealed there, and Anoni puts your real data back into the AI's reply. That is also why the connector writes markers rather than believable fake names. With no vault to go back through, a fake name would be impossible to tell apart from real data.