> ## Documentation Index
> Fetch the complete documentation index at: https://docs.dovetail.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Agent triggers

A trigger defines when or how your agent runs. The right choice depends on how the work actually happens—on a cadence, in response to something, or only when you ask.

## Trigger types

| Trigger          | When to use it                                                                                                                           |
| :--------------- | :--------------------------------------------------------------------------------------------------------------------------------------- |
| On-demand        | You want full control. The agent only runs when you click Run now or invoke it in chat.                                                  |
| Scheduled        | The work is recurring on a predictable cadence—daily digests, weekly summaries, monthly rollups.                                         |
| Dovetail events  | The work should happen the moment something changes inside Dovetail—a new project data point, a new highlight, a new doc, a tag applied. |
| External webhook | The work is triggered by a system outside Dovetail. See External webhook triggers.                                                       |
| Digital twin     | The agent should behave as a continuous active listener in chat. See Digital twins.                                                      |

## On-demand

The simplest trigger. No schedule, no events—the agent sits idle until you run it manually or mention it in chat with @.

Good fits: one-off analysis with deep skills docs, exploratory summaries, agents you invoke conversationally, agents you only need to use at specific times for a specific task.

## Scheduled

Scheduled agents run on a recurring cadence you define. Use them for anything predictable—a Monday morning digest, a Friday competitor scan, an hourly triage sweep.

### Choosing a cadence

Match the schedule to the freshness of the data. If new items land every few minutes, a daily digest is probably enough. If items land once a week, a daily run wastes context and produces empty output most days.

Common cadences and when they fit:

* **Hourly**—high-volume sources where fast response matters
* **Daily**—inbox-style summaries, tagging sweeps, standup summaries
* **Weekly**—team digests, retros, competitor scans, trend reports
* **Monthly**—executive rollups, quarterly prep

### Setting the schedule

Schedules use a picker with natural language input—describe when you want the agent to run and Dovetail translates it into a cadence. Examples: "every Monday at 9am," "weekdays at 5pm," "the first of every month."

### Timezone behavior

Schedules run in your local timezone by default. You can override the timezone in the schedular UI if the agent should fire for a different region or a shared team time.

## Dovetail events

Event triggers fire the moment something changes in your workspace. The agent runs immediately, with the changed item as its input.

### Available events

| Event                 | Fires when                                  |
| :-------------------- | :------------------------------------------ |
| Data added to project | A new data is added to a specific project   |
| Highlight created     | Someone creates a highlight on a transcript |
| Doc created           | A new doc is created in a specific folder   |
| Tag applied           | A specific tag is applied to any data       |

### Example patterns

**Auto-tag new project data**

> Trigger: Data added to project Instructions: When new data lands in the project, apply tags from the attached taxonomy. If the data doesn't clearly match any tag, leave it untagged rather than force a match.

**Notify on churn signals**

> Trigger: Tag applied (Churn risk) Instructions: When this tag is applied, summarize the data point in two sentences and post to the #cs-alerts channel in Slack.

**Enrich new docs**

> Trigger: Doc created Instructions: When a new doc is created in the \[Folder name] folder, generate a one-paragraph summary and add it as a comment at the top of the doc.

## External webhook

For triggers coming from Salesforce, Linear, Jira, or any external system, see External webhook triggers.

## Digital twin

Digital twin agents don't use the trigger types above. They run as continuous active chat listeners when invoked. See Digital twins.

## Choosing the right trigger

Ask two questions:

1. **Does the work happen on a cadence, or in response to something?** Cadence → Scheduled. Response → Dovetail event or External webhook.
2. **How fast does the response need to be?** Real-time → event. Same-day → schedule. Only when needed → on-demand.

When in doubt, start with on-demand. Once the agent is producing good output reliably, move it to the trigger that reflects the real workflow.