> ## Documentation Index
> Fetch the complete documentation index at: https://docs.superember.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Run history and approvals

> Understand how SuperEmber keeps workflow execution inspectable and human-governed.

Run history and configured approvals are the trust layer for SuperEmber workflows.

This page describes workflow-owned waits. Automatic tool review, delegated escalation, and Personal Agent approvals use different surfaces; see [Approvals, review, and inbox](/product/approvals-and-inbox).

They let teams see what happened, why a run paused, which agent or connector acted, and where human approval was required.

## Run history

A workflow run records the version of the workflow package that executed.

The run inspector should make these details visible:

* trigger and input
* step status
* agent tasks
* fleet delegation and joins
* connector actions
* waits and resumptions
* approvals and decisions
* artifacts and outbound messages
* errors, retries, and replay context
* version that executed
* governance decisions

## Typed waits

Workflows use typed waits to pause and resume long-running work.

Common wait types include:

| Wait type           | Meaning                                               |
| ------------------- | ----------------------------------------------------- |
| `delay`             | Resume after a time interval                          |
| `approval`          | Resume after a human approves or rejects              |
| `human_input`       | Resume after a person provides information            |
| `choice`            | Resume after a person selects an option               |
| `external_callback` | Resume after an external system calls back            |
| `channel_reply`     | Resume after a Slack or Teams reply                   |
| `agent_result`      | Resume after an async agent task completes            |
| `fleet_join`        | Resume after child agent work reaches the join policy |

## Approvals

Approvals keep humans in control of risky or externally visible actions.

A workflow can request approval before:

* sending an external message
* calling a connector action
* publishing or modifying a workflow
* continuing after a low-confidence agent result
* spending budget beyond a configured threshold

Approval events should be linked back to the run, the step, the user, the decision, and the resulting action.

## Replay

Replay is useful for debugging and trust.

Because runs pin the workflow version and preserve execution context, teams can understand what changed between an old run and a new run without relying on a hidden conversation transcript alone.

## Inbox

Open workflow inbox items represent runs that need a human decision or response.

Use the inbox for approval, input, choice, and channel-reply waits. Each item links back to the workflow and run context so the reviewer can decide with enough context.