Skip to main content

Tasks & Automations

Think of your agents as employees who can work two very different shifts. Sometimes you hand one of them a single file and say "take care of this one"; other times you set up a routine that runs on its own every morning. Tasks and Automations are those two shifts — both reuse the agents you've already built, but they differ in where the work comes from and how often it runs.

  • Agent Tasks — run one agent on a single record (one row of one of your databases), using that row's own data as context. You give it one instruction, it runs once, and it writes the result back. Best for one-off work that depends on the specific row in front of it.
  • Agent Automations — chain agents together into a multi-step routine that runs on a recurring trigger (the thing that kicks the routine off — a schedule, an incoming alert, an event, or a button you press). Best for work that needs to keep happening over and over, often across several services.

What you can do here

You can…What that means for you
Run an agent on one recordPick a row, write a one-line instruction, and the agent acts on just that row — summarize a lead, draft a reply, score an application
Use a row's own data as the agent's contextThe agent reads every field in the record before it does the work, so its answer fits that exact row
Trigger the next agent when a row changesOne agent can update another record, which sets off a second agent — work flows row by row without you watching
Run a routine on a scheduleHave agents do the same job every morning or every hour, with no one pressing a button
React to outside eventsStart a routine the moment a new form is submitted or an alert comes in from another service
Coordinate several agents in stepsLine up specialist agents — one researches, one writes, one sends — and let the output of each flow into the next

How it fits together

A task acts on one row using that row's data. An automation listens for a trigger and runs a chain of agents. The two work well together — a routine can add rows to a database, and those new rows can each kick off a task.


A concrete example

Imagine a sales pipeline where new leads arrive through a Facebook form:

LayerToolWhat it does
Where leads come inMeta Forms connectorNew form submissions land as records in the Leads database
Per-record workAgent Task on LeadsEach new row sets off a "Qualify this lead" task — the agent reads the lead's notes, scores them, and sets a Status
Recurring batch workAgent AutomationEvery morning a routine scans for high-score leads, writes personalized outreach copy, and sends it to Slack for the sales team

Two patterns, one set of agents.


What's next