Your First Run
Run your first AI agent in under 2 minutes — watch MCP tool calls, model reasoning, and outputs unfold in a live trace timeline inside Invoked.
Overview
A run is a single agent execution — a sequence of model calls and tool invocations that accomplish a goal. Runs are the core unit in Invoked: everything else (skills, evals, analytics) is built around them.
Starting a run
- Open the Runs panel from the left sidebar (or press
⌘1) - Click New run (or press
⌘N) - Type your prompt in the input field at the bottom
- Press Enter or click Run
The agent will begin executing, and you'll see the run unfold in real time — tool calls, model reasoning, and outputs are all visible as they happen.
Reading the run trace
Each step in the run trace shows:
- Tool name — which tool was called (e.g.,
read_file,web_search) - Input — the arguments the model passed to the tool
- Output — the tool's response
- Duration — how long the step took
- Token usage — prompt and completion tokens for model steps
Click any step to expand it and see the full input/output payload.
Run states
| State | Meaning |
|---|---|
| Running | Agent is actively executing |
| Complete | Agent finished successfully |
| Error | Agent hit an unrecoverable error |
| Cancelled | You stopped the run manually |
Stopping a run
Press Escape or click the Stop button in the run header to cancel an in-progress run. The partial trace is preserved.