Figaro · Security

Security

myfigaro.ai · last updated July 2026

This page explains how Figaro protects your data and what our security model looks like in practice.

1. Consent-gate model

Figaro operates on a consent-gate principle: nothing is sent, booked, posted, or paid without your explicit approval. Every action that touches the outside world (email send, calendar booking, social post, payment) requires either a pre-set rule you authored, or a real-time tap of approval from you.

You can inspect, modify, or revoke any pending action before it executes. The default posture is draft-and-wait, not fire-and-forget.

2. BYO Claude Max — no Figaro LLM middleman

Figaro does not run its own LLM inference layer between you and Anthropic. Your inference goes directly to your own Claude Max account via Anthropic's API, the same channel as opening Claude.ai in a browser.

Anthropic's API terms explicitly prohibit using API traffic to train models. Your conversations and data are not used to train Claude or any other model.

Because you bring your own Claude Max, Figaro never holds or processes your Claude credentials — they stay in your environment.

3. Credentials encrypted at rest

API keys, OAuth tokens, and other credentials you connect to Figaro are encrypted at rest using AES-256. On the cloud setup, they are stored in a secrets store with least-privilege access controls. On your own machine setup, they live on your hardware and never leave it.

4. No human reads your mail

Figaro employees and contractors do not have access to your email, calendar events, or message contents. Access to production data is restricted, audited, and requires multi-party approval for any break-glass scenario.

Figaro may process email metadata (sender, subject, timestamp) to route tasks correctly, but the content of your messages is processed only by your Claude Max account — not stored on Figaro's servers beyond the duration of an active task.

5. Responsible disclosure

If you discover a security vulnerability in Figaro, please report it to [email protected]. We aim to acknowledge reports within 48 hours and to resolve confirmed issues within 30 days.

Please do not publicly disclose vulnerabilities before we have had a chance to address them. We appreciate responsible disclosure and will credit researchers who report valid issues.


Questions? [email protected]