Agent 'Polly' for scheduling and email with sandboxed identity

Security-First OpenClaw Setup: Agent 'Polly' with Sandboxed Identity

Claire Vo, CEO of ChatPRD, built an OpenClaw agent named Polly for scheduling and email management. She took a security-first approach, creating a completely separate digital identity for the agent with sandboxed permissions.

Sandboxed Agent Identity with Separate Credentials and Cost Monitoring

Claire refused to give the bot direct access to her personal accounts. Instead, she created a new digital identity (polly.the.bot@...) with its own email, calendar access, and carefully limited permissions. The approach:

• Sandboxed identity: Polly has its own email address and credentials, separate from Claire's personal accounts
• Limited permissions: Only the access needed for scheduling and email tasks
• Cost monitoring: Claire tracked API costs carefully to understand the real expense
• 3 real workflows: Demonstrated scheduling, email triage, and follow-up management in 24 hours

Scheduling and Email Automation Without Exposing Personal Accounts

Scheduling and email automation with a security-first approach that other founders can replicate. Claire proved you don't need to hand over your personal credentials to get value from an AI agent.

Built by Claire Vo, CEO of ChatPRD

Claire Vo, CEO of ChatPRD. "I refused to give the bot direct access to my personal accounts." Published on ChatPRD's "How I AI" series.

I refused to give the bot direct access to my personal accounts.