Skip to main content
A run is one execution of an agent. Every run gets an ID starting with wr_, opens in Studio at /runs/<run id>, and is kept in Runs with its recording, screenshots and outputs.

​
Run an agent

1

Open the run form

In Studio, click Run in the top bar. From Agents → All Agents, click the play button (Create New Run) on the agent’s row. Both open the run form at /agents/<agent id>/run.
2

Fill in the inputs

Enter a value for each of the agent’s inputs. Inputs with a default are prefilled. If the agent has a Login block, a Login credentials section asks which saved credential to use.
3

Click Run agent

The run starts and Studio opens on it, with the Browser and Run panes.
Run form with Inputs, Settings (Webhook Callback URL, Proxy Location, Run With, AI Fallback), Advanced Settings, Copy API Command and Run agent The defaults work for most runs. The settings you are most likely to change: Advanced Settings holds browser options: Browser Type, Browser Session ID (run in an open session), Reuse browser session, Browser Profile (start with saved cookies and logins), Browser Address (connect to your own browser over CDP), Extra HTTP Headers, CDP Connect Headers and Max Screenshot Scrolls. Browsers explains sessions and profiles. Copy API Command copies the same run as a cURL or PowerShell command, with your inputs and settings filled in. Use it to trigger the agent from a script. The code quickstart covers the SDKs.
Run is disabled while the agent has no blocks. Add one, or ask Copilot to build the agent first.

​
Watch a run live

The Browser pane shows the browser the agent drives. Its header switches between three views:
  • Live: the browser right now, while the run is in progress.
  • Recording: the video of the run, available once it has recorded.
  • Screenshots: the screenshots Skyvern took at each step.
Run pane on Timeline while the run is Running, with the Stop button in the top bar

​
Take control of the browser

Click Take control in the live view to use the mouse and keyboard in the browser yourself. A You’re in control bar shows while you have it. Use it to solve a CAPTCHA, get past an unusual login, or close a pop-up the agent is stuck on. Click Stop controlling to hand the browser back.

​
Stop a run

Click Stop in the top bar, then Stop run to confirm. The run ends with status canceled.

​
Read a run

The Run pane has three views, plus more under ⋯: A completed run with the Browser pane on Screenshots and the Run pane Timeline showing the block and its action API & Webhooks in the Run pane header copies this run as a cURL or PowerShell command and offers Test Webhook. When a run fails, the top of the Run pane names the failing block and the reason. Fix with Copilot sends the failure to Copilot, which proposes a fix in the Copilot pane. Retry opens the run form to run it again.

​
Re-run

After a run ends, the top bar’s Run button reads Re-run. It opens the run form prefilled with that run’s inputs, so you can change a value and run again.

​
Run statuses

A block that never ran because an earlier branch or condition skipped it shows skipped.

​
Download recordings, screenshots and files

To pull artifacts from code, call GET /v1/runs/{run_id}/artifacts. Each artifact comes with a signed_url you can download from. Filter with artifact_type, for example recording or screenshot_final:
Signed URLs expire. Request a fresh one each time instead of storing it. The API reference lists every artifact type.

​
Find past runs

Click Runs in the sidebar to open Run History, every run in your organization, newest first. Run History with the search box, Filter by Agent, Filter by Type and Filter by Status, and the runs table
  • Search by run ID, input value or credential (“Search by run ID, input, credential…”).
  • Filter by agent, by run type (Task, Workflow, Agent (CUA)) and by status.
  • Columns are Run ID, Detail, Status, Created At and Duration (some organizations also see Created By). Runs started by a schedule carry a badge next to the ID, and runs that used generated code show a lightning icon.
  • Click a run to open it in Studio.
To see only one agent’s runs while you work on it, open the Past Runs list from the Studio top bar.

​
Schedule an agent

A schedule runs an agent automatically on a repeating timetable. New schedules are Active and fire at the next scheduled time.
1

Open Schedules

Click Agents → Schedules in the sidebar, then Create Schedule.
Schedules page with the search bar, Filter by Status, and the Create Schedule button in the top right corner
2

Pick the agent

Search for it in Agent. Only saved agents appear.
3

Set the timetable

Fill in the fields below, and check Next Scheduled Runs to confirm the times.
4

Create

Click Create Schedule.
Create Schedule form with Agent, Name, Description, Repeat, At, Timezone, Advanced (cron expression), and Next Scheduled Runs
Runs must be at least 5 minutes apart. Shorter intervals are rejected.

​
Manage schedules

Schedules lists every schedule in your organization with Agent, Name, Schedule, Next Run and Status (Active or Paused). Search by agent or schedule name, or use Filter by Status. Each row’s menu can Pause, Activate, Duplicate or Delete the schedule. Click a schedule to open its page at /schedules/<agent id>/<schedule id>:
  • The switch in the page header pauses or resumes the schedule, and the trash icon deletes it.
  • Schedule Configuration: frequency, timezone and cron. Click Edit to change the timetable, name or description.
  • Agent Inputs: the values the schedule passes, or the defaults it falls back on.
  • Details: Created and Last Modified.
  • Upcoming Runs: the next times it will fire.
Paused schedules keep their settings and start no runs. Deleting a schedule is permanent and leaves its past runs in place. To point a schedule at a different agent, create a new schedule. Scheduled runs appear in Run History with a scheduled badge.

​
Next