Skip to main content
Prints an agent’s turns: one block per task run, in run order. Each turn shows what triggered the run (your initial prompt, a follow-up, a PR comment, a PR review, a CI result), the agent’s written response, and an error if the run failed. Use it to read back what an agent did and why, without opening the dashboard. A turn is the unit the dashboard’s Turns view renders. murmur turns reads the same data from the command line.

Synopsis

The slug is optional — omit it to read the turns of the agent resolved from your current context (on a VM, the agent itself).

Arguments

Output

Each turn renders as a block: a header with the run number and its start and end times (UTC), the task description, each input with its kind and label, the agent’s response, and an error line if the run failed. An in-flight turn shows (in flight) for its end time.
With --json, the command prints a JSON array of Turn objects:

Examples

Read an agent’s turns

Read every execution of a slug

Includes turns from prior executions — useful after a --resurrect or --force-new spawn reused the slug.

Read the current agent’s turns on a VM

Errors