1. OAIZ records each pass of a workflow as a run. This video shows how to read runs in Logs.
Runs and logs

What runs and logs show

What each run did, and how to find out why one stopped.

Sort the request: damaged item, delivery question, return or refund.
Draft a reply that follows the Returns policy.
Send the approved reply to the customer.
LogsLive Feed
26 Oct 2026, 10:04:12Triggered runFailed Stopped at Draft a reply1 step completed
  1. Maya Chen emails about a cracked lamp shade. Her email is an event, and it starts the first job.
  2. Each job publishes an output event, and the jobs listening for it continue the run.
  3. Everything that one email started is one run. Jonas Berg's email starts a run of its own.

An event starts the first job

Each job passes an output event on

One event, one run

PublishedLive
Email received
Sort the request: damaged item, delivery question, return or refund.
Route by category.
Draft a reply that follows the Returns policy.
Sam approves the reply before it is sent.
Send the approved reply to the customer.
Tell #support about refund requests.
Cracked lamp shade · Maya Chen
Maya ChenCracked lamp shade, order 4821 One run
Jonas BergWhere is my order, order 4790 Its own run
  1. As a run reaches each job, the job shows a status. Succeeded means its action finished.
  2. Failed means the action returned an error.
  3. Filters not met means the event did not pass the job's Filters. That is not an error.
  4. Did not run means no event reached the job, for example because an earlier job failed.
  5. Waiting for approval means the run is holding until a person decides.
  6. Any failed job makes the run Failed. Paths that Filters stopped do not, so Succeeded does not mean every job ran.

Each job shows a status

Failed means an error

Filters not met is not an error

Did not run means no event arrived

Waiting for approval holds the run

One failed job fails the run

The action finished
Returned an error
Did not pass its Filters
Not an error
No event reached it
Waits for a person
A run Succeeded Failed Did not run Failed
A run Succeeded Succeeded Filters not met Not every job ran Succeeded
  1. Run workflow on a draft starts a test run. Write starts off, so it sends nothing.
  2. LLM calls still run in a test, so you read real answers.
  3. Once the workflow is live, each matching email starts a Triggered run that acts for real.

A draft runs as a test

LLM calls still run

Live runs act for real

You're in draft mode, only read permissions are on by default
PermissionsChoose what this run is allowed to do for real. Change event
ReadNothing here reads from your connected integrations.
Write Gmail · Create Draft Gmail · Send Email Slack · Send Message
Publish Events Each step publishes an event that other jobs and workflows can listen for.
Always runs This job executes on every run, including test runs Sort the requestcalls GPT-5.4 Mini
Run workflow Logs
Write is off
Gmail and Slack return example results. Nothing is sent.
LLM calls run in every test
Sort the request calls its model, so you read the real answer.
PublishedLive
Refund request · Lena Ortiz
Triggered run
Gmail · Create Draftcreates a real draft
Slack · Send Messageposts to #support
Approvalwaits for Sam
  1. Logs, beside Run workflow, lists each run with its status, a summary, the duration and the cost.
  2. Open Maya's run. Steps lists each job with its own status.
  3. Input & output shows the event a job received and the event it published.
  4. The Cost tab shows the job's cost, tokens and model calls.
  5. On the failed job, the Execution tab gives the error and the call that failed.

Every run is in Logs

A run lists its steps

What a job received and published

What each job cost

Why a job failed

Run workflow Logs
Logs Live Feed
All statuses Last 30 days
Started Status Summary Duration Cost
26 Oct 2026, 10:04:12Triggered runFailed Stopped at Draft a reply2 steps completed 3.104s€0.0012
26 Oct 2026, 09:52:30Test runSucceeded All steps completedOutput event emitted 6.204s€0.0021
25 Oct 2026, 16:18:45Test runSucceeded All steps completedOutput event emitted 5.718s€0.0019
Logs Run resultsFailedMon, 26 Oct 2026, 10:04:12
Steps
Sort the request10:04:12Succeeded
Route by category10:04:13Succeeded
Tell #support10:04:13Did not run
Draft a reply10:04:14Failed
Approve the reply10:04:14Did not run
Send the reply10:04:14Did not run
Select a job to inspect
Sort the request
ExecutionInput & outputCost
Input event
from: "maya.chen@example.com"
subject: "Cracked lamp shade"
text: "My lamp shade arrived cracked. Order 4821."
Output event
category: "damaged item"
order: "4821"
Cost & usage
Total cost€0.0011
Total tokens1,482
Model calls1
Token usage
Input1,310
Output172
Draft a reply
ExecutionInput & outputCost
Summary
StatusFailed
Execution time0.8s
CostNo cost recorded
Gmail · Create Draft - the recipient address is missingExplain in chat
Execution logs
Integration: GmailFailedCreate Draft
Tool errorThe recipient address is missing.
  1. Compare the events. Maya's address came in, but the first job did not pass it on.
  2. Explain in chat gives Chat the job, its log and its input, and asks why it failed.
  3. Ask Chat to fix the first job. Test the draft, then choose Go live.
  4. Then choose Re-run step on that job. It runs with Maya's original email and the live fix.
  5. A re-run acts for real, so the jobs after it run too, and this time the reply is drafted.

Find why it failed

Or let Chat explain it

Fix it, test it, then go live

Then re-run the first job

This time the reply is drafted

Sort the request · Input event
from: "maya.chen@example.com"
subject: "Cracked lamp shade"
text: "My lamp shade arrived cracked."
Sort the request · Output event
category: "damaged item"
order: "4821"
from: missing
Draft a reply · Failed
With no address in the event, Gmail · Create Draft has no recipient.
Gmail · Create Draft - the recipient address is missingExplain in chat
The jobDraft a reply
Its execution logEvery call, in order
Its input eventWhat it received
Chat starts a new conversation and asks why the job failed.
Ask Chat
Sort the request should pass on the customer's address.
Auto
Test the draft
Customer requestsDraft
Run workflowLogs
SucceededTest run
Go live
Go live PublishedLive
Sort the requestSucceeded Open in workflow Re-run step
Status:Succeeded
Time started:26 Oct 2026, 10:04:12
Time finished:26 Oct 2026, 10:04:13
Input event
from: "maya.chen@example.com"
subject: "Cracked lamp shade"
text: "My lamp shade arrived cracked. Order 4821."
Re-run 1 step?

This will re-trigger this step with their original input events. Integration commands (Slack, GitHub, Gmail, etc.) will be executed and output events will be published.

CancelRe-run step
Maya's original email
The input event the job first received
The live configuration
The job as it is now, with the fix
Run resultsWaiting for approvalMon, 26 Oct 2026, 11:02:37
Steps
Sort the requestRe-run11:02:37Succeeded
Route by category11:02:38Succeeded
Draft a reply11:02:39Succeeded
Approve the reply11:02:40Waiting for approval
Draft a reply
ExecutionInput & outputCost
Summary
StatusSucceeded
Execution time1.2s
CostNo cost recorded
  1. You can now read any run, see how each job went, and find where it stopped.
  2. The docs pages Runs and Logs cover every status and setting.

Read any run

What started it
One event, and every job it reached
Triggered runTest run
How each job went
A status on every job
Succeeded Failed Filters not met Did not run Waiting for approval
Where it stopped
The error, the events in and out, and the cost
Explain in chatRe-run step
Go deeper in the docs Runs Logs