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

# VibeOps: Your First Conversation

> Learn the fundamentals of conversational cloud operations — talk to AI agents using natural language to manage your infrastructure.

## What is VibeOps?

VibeOps is CloudThinker's conversational approach to cloud operations. Instead of clicking through dashboards or writing scripts, you talk to specialized AI agents using natural language. Describe what you need, and agents handle the rest.

***

## Why Conversational Operations?

Traditional cloud operations require deep tool expertise. To find underutilized EC2 instances, you'd run `aws ec2 describe-instances`, join it with CloudWatch metrics via a script, filter by utilization thresholds, and format the output. To check compliance, you'd navigate Security Hub, apply filters, export a CSV, and build a report. Every question requires knowing which tool, which command, and how to interpret the output.

VibeOps eliminates this barrier. You describe what you need in plain language — `@alex find EC2 instances with low utilization` — and the agent handles the tool orchestration, metric correlation, and output formatting automatically. This means:

* **Non-specialists can run cloud operations** — developers, product managers, and executives can ask cloud questions without CLI expertise
* **Complex multi-step analysis in one prompt** — ask questions that would normally require multiple tools and scripts
* **Faster operations** — hours of manual analysis compressed to seconds
* **Institutional knowledge stays accessible** — senior engineers encode expertise in prompts that anyone can use

The `@agent #tool [instruction]` syntax is the core of VibeOps — learn it once, and it works across every domain.

***

<Steps>
  <Step title="Open a Workspace">
    Log in to [CloudThinker](https://app.cloudthinker.io) and select your workspace. If you haven't set one up yet, follow the [Quick Start](/quickstart) guide first.

    Make sure you have at least one cloud connection configured — agents need access to your infrastructure to return real results.
  </Step>

  <Step title="Meet the Agents">
    CloudThinker has five specialized agents. Each is an expert in their domain:

    | Agent                          | Mention   | What They Do                                         | Enabled When                                                                     |
    | ------------------------------ | --------- | ---------------------------------------------------- | -------------------------------------------------------------------------------- |
    | [Anna](/guide/agents/anna)     | `@anna`   | Coordination across agents, strategy, reporting      | Always available (enabled by default)                                            |
    | [Alex](/guide/agents/alex)     | `@alex`   | Cloud infrastructure, cost optimization, performance | [AWS connection](/guide/connections/aws), GCP, or Azure connection added         |
    | [Oliver](/guide/agents/oliver) | `@oliver` | Security, compliance, vulnerability assessment       | AWS, GCP, or Azure connection added                                              |
    | [Tony](/guide/agents/tony)     | `@tony`   | Database performance, query optimization             | [PostgreSQL connection](/guide/connections/postgresql) or MySQL connection added |
    | [Kai](/guide/agents/kai)       | `@kai`    | Kubernetes, containers, cluster management           | [Kubernetes connection](/guide/connections/kubernetes) added                     |

    **@anna** is your starting point — she's available immediately and can coordinate work across all other agents. The specialist agents activate automatically as you add the relevant [connections](/guide/connections/overview).

    <Tip>
      Don't see an agent? Check your connections in **Settings** > **Connections**. Each agent needs its matching infrastructure connection to become available.
    </Tip>
  </Step>

  <Step title="Write Your First Prompt">
    In the chat interface, type a natural language request with an `@agent` mention:

    ```
    @alex show me a summary of my AWS resources
    ```

    Alex will scan your connected AWS accounts and return a resource overview — instance types, regions, services in use.
  </Step>

  <Step title="Add Tool Commands">
    Tool commands (`#tool`) tell agents what output format you want:

    | Command      | Output                                                  |
    | ------------ | ------------------------------------------------------- |
    | `#dashboard` | Interactive visualization with charts                   |
    | `#report`    | Detailed analysis document                              |
    | `#recommend` | Actionable recommendations                              |
    | `#alert`     | Set up monitoring [notifications](/guide/notifications) |
    | `#chart`     | Data visualization                                      |

    Try combining an agent with a tool:

    ```
    @alex #dashboard show me EC2 cost breakdown by instance type over the last 30 days
    ```
  </Step>

  <Step title="Refine and Drill Down">
    Start broad, then narrow based on results. This progressive refinement is how VibeOps works best:

    ```
    @alex analyze my AWS spending trends this quarter
    ```

    Then follow up:

    ```
    @alex drill into the RDS costs — which instances are over-provisioned?
    ```

    And take action:

    ```
    @alex #recommend right-sizing options for the over-provisioned RDS instances
    ```
  </Step>

  <Step title="Try Multi-Agent Collaboration">
    For complex tasks, combine agents in a single conversation:

    ```
    @anna coordinate a full infrastructure review:
    - @alex analyze cost optimization opportunities
    - @oliver audit security posture
    - @kai check Kubernetes resource utilization
    ```

    Anna orchestrates the other agents and consolidates their findings into a unified report.
  </Step>
</Steps>

***

## Common VibeOps Patterns

<Tabs>
  <Tab title="Cost Analysis">
    ```
    @alex #dashboard AWS cost trends for the last 90 days
    @alex identify unattached EBS volumes and unused Elastic IPs
    @alex #recommend reserved instance purchases for stable workloads
    ```
  </Tab>

  <Tab title="Security Audit">
    ```
    @oliver audit security groups for public access on database ports
    @oliver #report compliance status against CIS benchmarks
    @oliver check IAM policies for excessive permissions
    ```
  </Tab>

  <Tab title="Performance Check">
    ```
    @tony analyze slow queries on production PostgreSQL
    @tony #dashboard query performance with P95 latency trends
    @kai analyze pod resource utilization across all namespaces
    ```
  </Tab>
</Tabs>

***

## Tips

* **Be specific**: "Analyze EC2 costs in us-east-1 for the last 30 days" works better than "show me costs"
* **Use context modifiers**: Add timeframes, regions, thresholds, and scope to narrow results
* **Follow up naturally**: Agents remember the conversation context — you can drill down without repeating yourself
* **Combine tools**: Use `#dashboard` for visuals, then `#recommend` for action items on the same topic

***

## Next Step

<Card title="Code Review" icon="code-pull-request" href="/guide/tutorial/code-review">
  Set up AI-powered code reviews for your repositories
</Card>