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

# AI DevOps

> Use agent sessions to plan, draft, continue, and review infrastructure changes without leaving the Git-based delivery path.

# AI DevOps

AI DevOps in Cloudgeni is the agent-session workflow exposed in the app at `Agent Sessions`.

It is the shared execution surface behind:

* Direct agent chats
* Some remediation handoffs
* Cloud resource import handoffs
* Session resume and fork workflows

## What It Is Good For

Use an agent session when you want Cloudgeni to work with repository context rather than only show
scan output.

Typical use cases:

* Draft new IaC for a requested change
* Refactor or explain existing IaC
* Continue from a finding or remediation thread
* Import existing cloud resources into a repository structure

## How A Session Works

1. Open `Agent Sessions`
2. Pick the repository or repositories you want the agent to work in
3. Optionally attach cloud and documentation context
4. Submit the prompt
5. Watch the streamed conversation and tool output
6. Resume or fork the session if you want a second pass

The backend service is run-oriented. A single session can have multiple runs over time, which is
why resume and fork are first-class behaviors in the codebase rather than add-ons.

## Terminal Workflow

Use the [CLI](/cli/overview) when you want to work with the same agent sessions from a terminal.

```bash theme={null}
cloudgeni org use
cloudgeni agent
```

`cloudgeni agent` opens the active session for the selected organization. To send a single prompt
and stream the result:

```bash theme={null}
cloudgeni agent "Investigate the latest drift and propose the smallest safe change"
```

For automation or another coding agent:

```bash theme={null}
cloudgeni agent run "Draft the Terraform change" \
  --repo <repo>:<branch> \
  --cloud-integration <integration> \
  --watch
```

## What The Agent Can Use As Context

The current app supports a guided launch model with context selectors for:

* Repositories
* Cloud integrations
* Documentation integrations

That makes the session composer more useful than a plain prompt box. Use it.

## How Changes Land

Agent sessions are still Git workflows.

The useful output is usually:

* File changes in the selected repository
* A branch or pull request
* A follow-up review cycle

The product is not designed around the agent writing directly to production infrastructure.

## When Not To Start Here

If you only want an initial security signal, [Static Analysis](/features/static-analysis) or
[Cloud Compliance](/features/cloud-compliance) is usually a better first step.

Start with AI DevOps when you already know you want generated or edited repository changes.

## Related Workflows

<CardGroup cols={3}>
  <Card title="Scheduled runs" icon="calendar-clock" href="/features/scheduled-runs">
    Run recurring agent sessions from a workspace schedule.
  </Card>

  <Card title="Cloud Resource Import" icon="sitemap" href="/features/cloud-resource-import">
    Start from discovered resources and hand off into an agent session.
  </Card>

  <Card title="Pull Request Reviews" icon="git-alt" href="/features/pull-request-reviews">
    Use agents on incoming pull requests rather than only in interactive sessions.
  </Card>
</CardGroup>