Skip to content

Your data
stays yours

Roles, guardrails and a full log between your AI and your ad accounts.

  • Drafts by default
  • Guardrails in code
  • No model training

Where your
data goes

Roger returns the data you ask for to the AI assistant you connected (Claude, ChatGPT or another MCP client), where that assistant's terms and your plan settings apply, and shares it only with the sub-processors listed on /privacy/ (for example Anthropic when you use Ask Roger). Roger never trains models on it.

Who can
get in

  • OAuth 2.1

    Your AI gets a token you can revoke. No passwords shared.

  • Name and email only

    Each teammate signs in on their own.

  • Roles per website

    Admin, Editor, Viewer. Viewers can't change anything.

  • One-click revoke

    Cut off an AI or a teammate at once.

Nothing changes
without a check

  • Guardrails

    Budget caps, protected campaigns, no removals. Checked before every change.

  • Drafts by default

    Changes wait for a confirm. New campaigns start paused.

  • Append-only Activity

    Who, which AI, what changed, why. Kept 25 months. Never prompts.

Verified by
third parties

  • Google Limited Use API Services User Data Policy
  • CASA Tier 2 Security assessment for Google data
  • EU hosting App, database and files
  • No training Your data never trains AI models

Request the documents

Found a flaw?
Tell us

Report it privately. We fix it and keep you posted.

security@heyroger.ai

Other questions: email support