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

# Using Atlas in the Web App

> Start an investigation, follow and steer its agents, read the report, share it, and schedule recurring research in the Sixtyfour web app.

Run and manage investigations from [Atlas](https://app.sixtyfour.ai/atlas) in the Sixtyfour web app. Atlas appears in the sidebar once it is enabled for your organization. See [Atlas Overview](https://docs.sixtyfour.ai/atlas/overview.md) for access.

## Start an investigation

1. Open [Atlas](https://app.sixtyfour.ai/atlas).
2. Under **New investigation**, enter a name, company or topic to research.
3. Optionally, attach files and set the options below.
4. Press Enter to start. Shift+Enter adds a line.

| Option | What it does |
| - | - |
| Engine | XHigh, or Scout when it is enabled for your organization. |
| Budget | XHigh runs on 1×, 2× or 4× the price of one unit, or a custom amount up to 10×. Scout runs are always one unit. |
| Time limit | No limit, 15 minutes, 30 minutes, 1 hour, or a custom limit from 10 to 1,440 minutes. When time is up, the agent writes up what it found. |
| Private | Only you can see the run, while you're on its team. Appears when your team allows private runs. If you leave the team or the organization, its owners can return the run to the team or give it to someone on it, a week after you lost access; you're told by email when that happens. |
| Use my saved instructions | Applies your [saved instructions](#save-instructions). Appears once you have saved some, and is on by default. |

You can attach up to 5 files: PNG, JPG, WebP or GIF images up to 20 MB each, and PDF, DOCX, XLSX or CSV documents up to 25 MB each.

### Start one investigation per row

1. Select **Upload CSV**. The file needs a header row, and each column becomes a field on the lead.
2. Check the rows in **CSV Preview** and add an optional research direction.
3. Select **Submit All**, which shows the total credits.

Each row starts its own investigation, up to 1,500 rows. Attached files aren't available on a CSV upload.

## Follow a run

An investigation opens on its graph while it runs, and on its report once it completes.

| View | What it shows |
| - | - |
| Report | The intelligence report, with citations that open the finding behind them. |
| Graph | Findings and the links between them. Search, or filter by nodes, types and confidence. Select a node to see how the agents reached it and the evidence for it. |
| Sources | Every cited page. Open one, copy its link or download it as a PDF. |
| Agents | The research team, and each agent's timeline. |
| Files | The notes and attachments in the workspace. |

The investigation's status shows how the run is going:

| Status | Meaning |
| - | - |
| Not started yet | The investigation exists and no run has begun. |
| Research in progress | Agents are working. |
| Stalled — no recent agent activity | The agents have gone quiet. |
| Research complete | The run finished. |
| Run stopped before finishing | The run failed or was stopped. Its findings are kept. |

The status covers the whole investigation, not one run. Once a run has completed, it reads **Research complete** between runs, and stays that way when a later run fails or is stopped.

A blue dot in the sidebar marks a report you haven't opened yet.

## Steer a run

You can change what a live agent works on without restarting it. This needs edit access to the investigation.

1. Open **Agents** and select the main agent or a subagent.
2. Type in **Send a message…** and send it. Messages can be up to 4,000 characters, and queued messages appear in the timeline until the agent reads them.

To point the agents at a note, use **Send to running agent** on that file.

To stop, use **Stop latest run**. Findings are kept, and you can add an agent later to carry on.

## Add more research

**Add Agent** starts another run on the same workspace and graph, during a run or after it has finished. It needs edit access and is billed as a new run.

1. Select **Add Agent**.
2. Enter a research direction. Later agents keep it.
3. Optionally, attach files and set the engine, budget and time limit.
4. Select **Deploy**.

To follow up on one finding, select its node in the graph and use **Dig deeper on this node…**, which also starts a billed run.

## Edit notes and findings

Notes save as you type. In the file list you can create notes and folders, and rename, move, duplicate and delete them, with undo. **File history** shows what changed between versions.

In the graph, **Add finding** records something you know, and you can edit a node, connect it to another node or remove a connection.

If a finding is wrong, use **Retire node**. The report leaves it out and later research won't add it back. **Restore node** undoes it.

## Read and revise the report

Atlas writes the report when a run completes. To read one sooner, use **Generate early report** while research is running, and **Update report** to bring it up to date.

To change a finished report, use **Revise report**. Add feedback of up to 4,000 characters and up to 5 evidence files, then select **Rewrite report**. Later reports keep your feedback. The person who created the investigation can revise its report once research has stopped. For an investigation started by an API key or a workflow, anyone with edit access can.

**Earlier reports** lists previous versions as PDFs.

## Share an investigation

Everyone on the team that owns an investigation can see it, unless the run is private. To share it further, use **Share**.

| With | How | What they can do |
| - | - | - |
| Another team | **Team access…**, as a team owner or org admin | **Can view**: open and export. **Can edit**: also add agents and edit, with their runs billed to their team. |
| Anyone with the link | **Public link** | Read the report, graph, sources and files, and download the PDF. |

A public link shows the investigation as it is now, so later research appears there. For organizations with customer-managed encryption, it shows a public copy frozen when it was published, without the PDF download; **Sync public copy…** republishes it from the workspace as it is now. Public pages hide the agents and file history, and search engines don't index them. A private run can't be shared.

To group investigations, add them to folders. **Share folder…** shares a folder with another team, including anything added to it later.

## Export

**Share** → **Export** downloads one investigation as a **Workspace ZIP**, a **Report PDF**, a **CSV**, or an **Identifiers CSV** that lists every email, phone number, IP address, name, organization, account and domain it found, one row each.

To export many, open [bulk export](https://app.sixtyfour.ai/atlas/export), pick investigations by search or date, and choose a format. A bulk export covers up to 5,000 investigations. ZIP and PDF exports build in the background and stay in **Recent exports** for 24 hours.

## Schedule recurring research

An investigation is a snapshot. An automation re-runs research on a schedule, so you catch what changes without starting over.

Open [Automations](https://app.sixtyfour.ai/atlas/automations), create one or start from an example, and fill in:

| Field | What it does |
| - | - |
| Name | What the automation is called. |
| Schedule | Every hour, day, week or month, at the time and timezone you choose. |
| What should it do? | **Re-deploy existing** adds a run to an investigation. **Create new** starts a new investigation each time, so each one has its own full report. |
| Research direction | What each run should look for. Optional. |

Automation runs use XHigh on a one-unit budget, billed as a normal run. A run fails if your balance is below the price of one run, or if the investigation already has a run going. An automation pauses if the person who created it loses access to its team. You can automate your own private investigations: only you see those automations, and one pauses if you leave the investigation's team. Automations can't target someone else's private investigation or one shared from another team.

## Save instructions

Saved instructions apply to every run, so you don't repeat them. Set them in [Settings → Agents → Instructions](https://app.sixtyfour.ai/settings/agents/instructions), up to 5,000 characters at each level:

| Level | Applies to | Who can edit |
| - | - | - |
| Everyone in your organization | Every run in the organization, including runs started with an API key | Org admins and owners |
| Just you | Runs you start, unless you turn them off for a run | You |

When both are set, a run follows both.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.