Skip to main content

​
From the UI

Open a prompt or workflow. The run form sits at the bottom right of the screen with input fields for each placeholder you’ve defined. Fill them in, hit Run, and the result appears in the right pane. A small key icon next to the inputs lets you pass runtime environment variables for the run – useful for populating ${VAR} placeholders in tool configurations. The run header shows duration, token usage, and cost at a glance. Click it to expand the details: token breakdown (input, cached, output, reasoning) and both gross duration (total processing time) and net duration (wall clock time from start to finish). For workflows, the canvas decorates each node with a status badge. Click any node to inspect its inputs and outputs. If a node ran multiple times (after a Fanout), a stepper lets you navigate between individual runs.

​
Threads and continuation

When running a prompt or workflow with memory, you can continue an existing thread or start fresh. The buttons appear at the bottom right of the run form once a previous run is loaded. Forking lets you branch a thread at any point: load an existing conversation, pick a message, and re-run from there with a different prompt revision. The original thread stays untouched. Invaluable for testing “would my new version have handled this better?”

​
Channels

When a workflow contains prompts on different channels, the run pane shows a tab for each channel. The Overview tab displays the workflow’s top-level inputs and outputs.

​
From the API

Trigger a run via POST to the API, store the run ID, wait for the webhook, then fetch the result. See the Integration Guide for the full flow.

​
Run history

The Runs button in the toolbar opens a scrollable log of every past run. Dates, revision numbers, tags, inputs, outputs, status, stats – all in one grid. Click any row to load that run and its revision for inspection.