xiaYuTian11/maskit

What is it?

Key features

• 01Masks API keys, connection strings, private IPs, phone numbers and other sensitive values before requests leave the machine.
• 02Restores placeholders in streamed SSE responses so output appears as normal typing.
• 03Keeps placeholder mapping consistent across a long multi-turn conversation.
• 04Works through local ports for Cursor, Claude Code, Codex and other tools with configurable Base URL.
• 05Browser extension covers ChatGPT, Claude and DeepSeek web pages, with more sites configurable.
• 06Provides a web console with logs, custom word lists, audit and token cost statistics.

Use cases

• Keeping credentials and internal hostnames out of prompts when using AI coding assistants.
• Sending internal documents to web-based AI chats without exposing personal data.
• Putting a privacy layer in front of a team's shared LLM relay or API gateway on a server.
• Testing which sensitive values get masked before deciding on a workflow.

Quick start

1. 1

   Download the desktop installer for Windows or macOS from the GitHub Releases page

2. 2

   Or run the Docker gateway with the one-line command from the README

   docker run -d --name maskit --restart unless-stopped -p 127.0.0.1:5801:5801 -p 127.0.0.1:18701:18701 -v maskit_data:/data -e MASKIT_PANEL_TOKEN="YourSecretToken123456" ghcr.io/xiayutian11/maskit:latest

Things to know

• !Desktop client supports only Windows and macOS; Linux users must use Docker.
• !Running from source requires manually placing the NER model files, which are not included in the repository.
• !Licensed under AGPL-3.0, a copyleft license that requires sharing source for networked modifications.

In collections

Topics

FAQ

What is maskit?

A local privacy gateway that masks sensitive data in prompts sent to LLM APIs and restores the original values in streamed responses. Maskit sits between AI coding tools or browser chats and external model providers. Requests are scanned locally for API keys, database connection strings, private IPs, phone numbers, ID numbers and similar data, which are replaced with structured placeholders before leaving the machine. When the model replies, the placeholders are restored to the original values in the streamed output, and the same placeholder is reused across a multi-turn conversation.

Is maskit free to use? What license does it have?

maskit is open source under the AGPL-3.0 license. Check the license terms before using it in commercial products.

Is maskit still maintained?

Actively maintained. The last commit to maskit was on Oct 21, 2026.

What can I use maskit for?

Keeping credentials and internal hostnames out of prompts when using AI coding assistants. Sending internal documents to web-based AI chats without exposing personal data. Putting a privacy layer in front of a team's shared LLM relay or API gateway on a server. Testing which sensitive values get masked before deciding on a workflow.

How do I install maskit?

Download the desktop installer for Windows or macOS from the GitHub Releases page Or run the Docker gateway with the one-line command from the README: docker run -d --name maskit --restart unless-stopped -p 127.0.0.1:5801:5801 -p 127.0.0.1:18701:18701 -v maskit_data:/data -e MASKIT_PANEL_TOKEN="YourSecretToken123456" ghcr.io/xiayutian11/maskit:latest

Where can I get maskit?

The source code and installation instructions are on GitHub: https://github.com/xiaYuTian11/maskit