Auth Your Agent is a security and access management tool that sits between AI agents and the websites they interact with. Its primary function is to allow users to approve, monitor, and revoke the actions that AI agents can perform on their behalf, without requiring the agent to hold the user’s passwords or long-lived API keys. The service operates through a phone-based approval flow: when an agent requests access to a site, the user receives a prompt on their phone to approve or deny the request, specifying the exact permissions granted. The user can revoke access to a single site or the entire agent with one tap. The tool offers two main modes of operation. For websites that have adopted the Auth Your Agent SDK, the site can verify the agent’s identity, the person it acts for, and the specific permissions granted, with sensitive actions such as payments or applications requiring a fresh, one-time approval on the user’s phone. For sites that have not adopted the SDK, the service uses a “take over” mode: the agent’s browser runs in a secure vault on the user’s machine, the user signs in manually on their phone, and the agent continues on the logged-in page without ever holding the session cookies. The vault signs the agent out when the task is complete. Optional integration with password managers like Bitwarden or Vaultwarden allows the vault to type saved credentials without exposing them to the agent. Typical use cases include delegating tasks such as booking appointments, making purchases, or filling out applications to AI agents while maintaining control over credentials and limiting the scope of each action. The service also provides a reporting mechanism: if an agent encounters a login form, CAPTCHA, or other obstacle it cannot handle, it can automatically file a report through its MCP server, which the owner can review. These reports are used to improve site coverage. Auth Your Agent is free during its early access period. It works in any modern browser and is also available as an Android app. The service provides separate guides for four user types: end users, agent builders, website operators, and AI agents themselves. Technical details regarding security and the underlying approval flow are documented in a published security whitepaper.