AgentGuard

Privacy Policy

Last Updated March 8, 2026

"We believe that the future of agentic commerce must be built on a foundation of absolute privacy and user sovereignty. This policy outlines our commitment to that future."

1. Data Sovereignty

Your data belongs to you. AgentGuard acts as a zero-trust middleware. When you connect an agent to services like Gmail or Slack, we do not store your raw credentials. Instead, we use encrypted Agent Passports (tokens) that are only decrypted at the moment of mission execution.

2. Information We Collect

We collect basic account information (email via Google Auth) and agent mission configurations. This includes the 'Custom Instructions' you provide, used exclusively to guide the reasoning engine of your specific agent.

3. Service Integrations (MCP)

When using Model Context Protocol (MCP) skills, AgentGuard strictly adheres to the 'Principle of Least Privilege'. If you authorize Gmail access, the agent can only read/send emails within the specific scope required for your described mission.

4. Data Sharing & Third Parties

AgentGuard does not sell or lease your data. Mission-critical data is sent securely to your selected AI Providers (e.g., OpenAI, Anthropic) for reasoning. This data is transmitted via encrypted channels (TLS 1.3) and is subject to the provider's own enterprise data privacy standards.

Questions about our privacy protocols? Reach out to our security team.

privacy@pootonline.com