Inspect a run in flight
Where to watch a run as it happens, and where every detail is saved.
When Vibestrate is doing work for you, you can watch it as it goes. There are three places to look.
Pick the surface that fits the moment. The terminal for a quick glance while it runs, the dashboard for the full live picture, and the files on disk for the complete record you can read back at any time.
The terminal A per-stage header right where you started the run. Status, the agent at work, and the check output streaming to your screen.
The dashboard Mission Control, a web view next to the run. A live timeline of every step, plus metrics, changed files, and the running diff.
The files on disk
Everything written under .vibestrate/runs/<runId>/. The full, append-only record of the run, there to read back whenever you want.
The terminal
If you start a run with the plain vibe run command, the terminal prints a header for each stage. The header shows the current status, the name of the agent doing the work, and any output it captured. When the checks run (the “validation” step that runs commands to confirm the work holds up), the output of those commands streams straight to your screen.
The dashboard
vibe run "..." --ui opens Mission Control, a web dashboard, next to the run. Submit a brief from Mission Control and it takes you straight to the run screen.
The main thing to watch on the run screen is the Live timeline. It shows one row per step in the flow. Each row gives you:
- its status,
- the role and profile in the seat (the “seat” is which kind of agent is in the chair and which settings it is using),
- the elapsed time, ticking up while it works,
- a live tail of what the model is writing right now.
Expand a row to see everything about that seat: the prompt it received, its full live transcript (the text, the thinking, and the tool activity, all as they stream in), and its response once it is done.
Around the timeline you also get:
- Live metrics - token usage and cost, as the provider reports them.
- Changed files - the files the run has touched so far. Click a file to see its diff or its full contents inline, read from the run’s worktree (the isolated copy of your project the run works in).
- Live execution - a raw console over every recorded provider stream.
The Git tab gives you the live diff against main as the run works, so you can see exactly what is changing.
The files on disk
Everything is recorded at .vibestrate/runs/<runId>/:
.vibestrate/runs/abc123/
state.json current status, transitions
events.jsonl every transition event, append-only
metrics.json tokens, durations, costs (where reported)
plan.md planner's output
architecture.md architect's output
execution.log executor's stream
validation.json commands run + exit codes + output
review.md reviewer's findings + decision
verification.md verifier's summary + decision
prompts/ raw prompts sent to each provider
outputs/ raw responses receivedevents.jsonl is the source of truth. Every state transition is one JSON line, and the file is append-only, so lines are only added, never changed. It is the record to trust when you want to know exactly what happened.
Use it to dig into a run after the fact:
cat .vibestrate/runs/abc123/events.jsonl | jq -c 'select(.type == "status-change")'Read past runs
To look back at a run that already finished:
vibe replay <runId>Replay is a read-only inspector for any saved run. It is handy for runs that finished long ago, runs from another machine that synced over, or any run you did not watch live.
Related
- Run state - what the status field means.
- Debug a failed run.
