Skip to content
AgentIDDocs

Type at least two characters to search titles, headings and content.

↑ ↓ to navigate↵ to open
AgentID Docs

Documentation

AgentID documentation

Govern how your organization uses AI and keep personal data and secrets out of AI providers.

AgentID checks every prompt before it reaches an AI provider. It masks personal data and secrets with typed placeholders, blocks prompts that break your policy, logs what happened for audit, and can restore the original values for the user on their own device.

You can protect AI in three places:

  • In the browser: ChatGPT, Claude, Gemini and Microsoft Copilot through the AgentID browser extension.
  • In Claude: Cowork, Claude Code and Claude Desktop chat through the AgentID plugin and the desktop companion.
  • In your own software: any LLM call in your backend, API or agent through the Node.js, Vercel AI SDK and Python SDKs, or the HTTP API.
Quickstart

Register an AI system and protect your first prompt in a few minutes.

How AgentID works

The guard check, masking, restoration and the activity log in one picture.

Browser extension

Protect ChatGPT, Claude, Gemini and Copilot in Chrome and Edge.

Claude

Protect Cowork, Claude Code and Claude Desktop chat.

SDKs and API

Guard and log LLM calls in your own application.

Data protection

What AgentID detects, how placeholders work and what is stored.

What's new

Browser extension 0.1.12 checks the whole message and every text file before upload, and detection is faster and more accurate. See the 2026-10-14 update. The 2026-10-14 release brought browser extension 0.1.11 with restoration across the whole conversation, one Claude Desktop integration for Windows and macOS, a Masked values view with audited Reveal in the activity log, and broader detection of personal data. See the release notes.