Skip to Content
SafetySafety

Safety

You’re always in control of what your assistant can do. Town provides multiple layers of safety controls to ensure workflows behave as expected.

Safety controls

The basics

Modes

Every workflow has a mode that controls execution:

ModeBehavior
AutonomousActions execute immediately
Approval-requiredWrite actions wait for your approval
Read-onlyCan only read, no modifications

Per-tool controls

You can set different modes for individual tools. A workflow might be autonomous for most actions but require approval for sending emails.

Action logging

Every action is logged. You can review what happened, when, and why — and adjust settings if needed.

Progressive trust

A common pattern is to start restrictive and loosen over time:

  1. Start with Read-Only — Workflow can analyze but not act
  2. Move to Approval-Required — Workflow can propose actions for your review
  3. Autonomous — When behavior is stable and low-risk

Default to approval-required for new workflows. Move to autonomous after you’ve verified behavior across multiple runs.

Built-in protections

Town includes security measures that apply automatically:

  • Blocked tool combinations — Certain combinations aren’t allowed for security
  • Prompt sanitization — Protects against prompt injection
  • Output validation — Ensures actions match expected patterns
  • Account isolation — Workflows can only access accounts you’ve connected

Learn more in Security.

Best practices

Start narrow, expand carefully. Begin with minimal tools and permissions. Add more as you verify behavior.

Review run logs regularly. Even autonomous workflows should be reviewed periodically to ensure they’re behaving correctly.

Use approval-required for email sends. Sent emails can’t be unsent. Review before sending.

Test before deploying. Use dry runs to verify workflow behavior without making real changes.

Last updated on