---
title: Runs
description: Start a run, follow the agents as they work, and find past runs.
---

A run is one piece of work handed to a factory. You can start it on the web, by mentioning the factory in Slack, from a Linear issue, or with an [automation](/automations). Every run shows up under **Runs**, wherever it started.

## Start a run

1. Select **New run** in the sidebar.
2. Pick the factory the run works in. A run stays in that factory, and you need at least one factory to send it.
3. Describe what you want and press Enter. Shift+Enter adds a new line.

The starting points under the message box (**Fix a bug**, **Build a feature**, **Add tests** and **Explain the code**) fill in the opening words for you.

### Attach files

Attach up to 4 files per message with the **+** button, by dragging them onto the message box, or by pasting them.

| File type                    | Size limit |
| ---------------------------- | ---------- |
| GIF, JPEG, PNG or WebP image | 5 MB       |
| UTF-8 text file              | 256 KB     |

## Follow a run

The run page updates live as the agents work:

- **Working** groups the steps of each turn (files read and edited, commands run, searches) under a timer, and becomes **Worked for …** when the turn ends. Failed and denied steps are marked.
- **Thinking** rows hold the agent's reasoning. Expand one to read it.
- The reply appears after the steps.

Send follow-ups from the message box at the bottom of the page. Each prompt shows who sent it, including prompts sent from Slack or Linear.

The run header has these controls:

- **Stop run** (the square button) stops the agent.
- **Run options** (⋮) turns **Show thinking** and **Show tool calls** on or off. Your browser remembers the choice for every run. It also links to **Runs from this factory** and **Factory settings**.
- **Open in Slack** or **Open in Linear** appears on runs that started there.

## Run statuses

| Status    | Meaning                                                     |
| --------- | ----------------------------------------------------------- |
| Running   | An agent is working.                                        |
| Complete  | The latest turn finished.                                   |
| Failed    | The run hit an error. The reason shows in the runs list.    |
| Blocked   | The run is waiting for an approval, an answer or a sign-in. |
| Cancelled | Someone stopped the run.                                    |

## Find a run

**Runs** lists every run in the workspace, 100 to a page. You can:

- filter by status with the tabs at the top
- search run titles
- filter by **Source** (where the run started: **AINOX Web**, **Slack**, **Linear** or **Automation**) and by **Factory**
- sort by **Newest first**, **Oldest first** or **Title A–Z**

The list refreshes itself while runs are in progress.

## Inbox

**Inbox** lists blocked runs from the whole workspace: runs waiting for an approval, an answer or a sign-in. The count next to **Inbox** in the sidebar shows how many are waiting.
