Operations agent overview

An operations agent does not talk to people. It runs a flow of actions when something happens: a call ended on your PBX, a webhook arrived, a schedule fired, another agent handed over data, or you pressed Run. Typical jobs: transcribe every call and email a summary, answer the same questions about every call in an Excel report, post each transcript to your CRM, archive recordings to your own storage, alert the team about missed calls.

You draw the flow on a board: a trigger, then steps, connected by arrows. Several arrows out of one step run in parallel; a step with several arrows in waits for all of them. Every run is recorded step by step, with what each step received and produced, so you can always see what happened and why.

The agent's pages#

Opening an operations agent shows its Overview: whether it is on, what starts it, the next scheduled run, the last run, the recent runs and this month's cost. The menu on the left leads to:

• Workflow: the designer, the board of triggers and steps.
• Runs: every run with its steps, what each received and produced, and the cost.
• Logs: every step of the last 100 runs on one line, filterable to what failed or waits for a retry.
• Settings: name, description, on or off, the webhook address, deletion.

Create one in five minutes#

1. On Agents, click New agent and choose Operations agent. Give it a name; the designer opens.
2. Start from a template (the fastest way: 18 ready boards for Yeastar PBX, HubSpot and general use, each tested end to end), describe what you want to Design with AI, or add a trigger from the palette on the left yourself.
3. Add steps and connect them by dragging from a step's right handle to the next step's left handle. Click a step to set it up in the inspector on the right.
4. Save, then Test run with sample trigger data, and open the run to check every step.
5. Enable the agent. From now on its trigger starts it on its own.

Passing data between steps#

Any setting can use data from earlier steps with {{step.field}}, for example {{transcribe.text}}, {{summarize.text}} or {{trigger.caller_number}}. The inspector lists what each step produces and the values you can reference, so you rarely type a name by hand. A value that is not there when the run reaches the step is left empty.

Where to go next#

• Triggers: what can start a run, and how to capture a real event with Listen.
• Steps: every action you can put on the board and its settings.
• Storage (S3): files and folders on Amazon S3 or any S3-compatible service.
• Designing and testing: templates, Design with AI, test runs, testing a single step, history and export.
• Email templates: the branded layouts the Send email step can use.
• Runs, retries and reports: what a run records, the retry policy, the runs report and the Excel export.
• Pricing: how runs are billed in automation minutes.