Modes
Modes control how much autonomy your assistant has. Choose based on the risk level and how much you trust a workflow’s behavior.
The three modes
Autonomous
The workflow executes all actions immediately without asking for approval.
What happens:
- Reading actions run immediately
- Actions that make changes (labeling, archiving, sending) also run immediately
- No approval requests are sent
- Fastest execution
Best for:
- Well-tested, trusted workflows
- Low-risk actions (labeling, archiving)
- High-frequency triggers (like every incoming email)
- Workflows that only report information to you
Autonomous mode can affect many emails quickly. Use it for low-risk actions and workflows you’ve already tested.
Approval-required
The workflow proposes actions that make changes, then waits for your approval before executing them.
What happens:
- Reading actions run immediately
- Actions that make changes wait for your approval
- You receive notifications with approve/reject options
- Actions only execute after you approve them
Best for:
- New or untested workflows
- Workflows that send emails
- High-stakes automations
- Workflows that access sensitive tools
Read-only
The workflow can only read and analyze — it cannot make any changes.
What happens:
- Reading actions run normally
- Actions that make changes are blocked entirely
- Cannot label, archive, or send
- Can still send reports to you via the “send email to you” tool
Even in read-only mode, your assistant can still send you emails with results and summaries. It just can’t modify your inbox or connected apps.
Best for:
- Analytics and reporting workflows
- Research workflows
- Testing new configurations
- Workflows that only gather and summarize information
Which actions require approval?
In approval-required mode, actions that make changes require your approval:
| Action | Needs approval? |
|---|---|
| Reading an email | No |
| Searching emails | No |
| Viewing your calendar | No |
| Adding a label | Yes |
| Archiving an email | Yes |
| Sending an email | Yes |
| Creating a draft | Yes |
| Deleting or trashing | Yes |
| Creating a calendar event | Yes |
| Modifying a GitHub file | Yes |
Per-tool approval overrides
You can set different modes for specific tools within a single workflow. This is one of the most powerful features of Town’s safety model.
Example: You want a workflow that automatically labels and archives emails (autonomous) but always asks before sending an email. You would:
- Set the workflow’s default mode to Autonomous
- Override the “send email” tool to Approval-required
Now labeling and archiving happen instantly, but sending always waits for your OK.
Choosing the right mode
| Situation | Recommended mode |
|---|---|
| Brand new workflow | Approval-required or Read-only |
| Workflow that sends emails | Approval-required (at least initially) |
| Simple labeling and archiving | Autonomous (once tested) |
| Research and analysis only | Read-only |
| Processing every incoming email | Autonomous (for speed) |
| Weekly summary workflow | Approval-required (review before sending) |
| Mission-critical automation | Approval-required (even when trusted) |
Progressive trust
A common pattern is to start restrictive and loosen over time:
- Start with Read-Only — See what the workflow would do without any risk
- Move to Approval-Required — Let it propose actions for your review
- Promote to Autonomous — When you’re confident the behavior is correct and low-risk
Best practices
Default to approval-required for new workflows. Move to autonomous only after reviewing several runs.
Keep email-sending workflows on approval-required. Sent emails can’t be unsent. Always review before sending.
Use read-only for experimentation. When testing new instructions or configurations, read-only mode prevents any accidental changes.
Check your approvals regularly. Approval-required mode only works if you actually review the requests. Don’t let them pile up.