> ## 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.

# AI Agents

> Five specialized AI agents for cloud operations — each an expert in their domain, coordinating seamlessly on complex tasks

CloudThinker's five AI agents replace the fragmented stack of cloud management tools with specialized experts that communicate, coordinate, and act on your behalf. Each agent has deep domain expertise, persistent memory of your environment, and access to the tools it needs to get work done.

<Note>
  Want to change tone, restrict access, or build your own agent? See [Customize Agents](/guide/agents/customize).
</Note>

***

## The Agent Team

| Agent                              | Role                  | When to Use                                                                             |
| ---------------------------------- | --------------------- | --------------------------------------------------------------------------------------- |
| [**Alex**](/guide/agents/alex)     | Cloud Engineer        | Cost analysis, infrastructure optimization, multi-cloud architecture, reserved capacity |
| [**Oliver**](/guide/agents/oliver) | Security Professional | Compliance audits, vulnerability assessment, IAM reviews, threat detection              |
| [**Tony**](/guide/agents/tony)     | Database Engineer     | Query optimization, performance tuning, index recommendations, capacity planning        |
| [**Kai**](/guide/agents/kai)       | Kubernetes Engineer   | Cluster health, pod right-sizing, autoscaling, RBAC audits, troubleshooting             |
| [**Anna**](/guide/agents/anna)     | General Manager       | Multi-agent coordination, cross-domain projects, executive reporting                    |

***

## Choosing the Right Agent

<Tabs>
  <Tab title="By Domain">
    | Need                                | Use       |
    | ----------------------------------- | --------- |
    | Cloud cost is too high              | `@alex`   |
    | Security audit or compliance report | `@oliver` |
    | Database is slow                    | `@tony`   |
    | Kubernetes cluster issues           | `@kai`    |
    | Problem spans multiple domains      | `@anna`   |
    | Don't know where to start           | `@anna`   |
  </Tab>

  <Tab title="By Output">
    | Output Needed                        | Use                |
    | ------------------------------------ | ------------------ |
    | Cost dashboard or spending breakdown | `@alex #dashboard` |
    | Compliance evidence report           | `@oliver #report`  |
    | Slow query recommendations           | `@tony #recommend` |
    | K8s OOMKill alert                    | `@kai #alert`      |
    | Executive quarterly review           | `@anna #report`    |
  </Tab>

  <Tab title="By Task Type">
    | Task                     | Use                                        |
    | ------------------------ | ------------------------------------------ |
    | One-time analysis        | Specific agent directly                    |
    | Scheduled recurring scan | `@alex #schedule` or `@oliver #schedule`   |
    | Cross-domain project     | `@anna` to coordinate                      |
    | Incident investigation   | `@anna` to dispatch all agents in parallel |
  </Tab>
</Tabs>

***

## Quick Reference: Prompt Patterns

```bash theme={null}
# Single agent — cost analysis
@alex identify EC2 instances with <20% CPU over 30 days

# Single agent — security
@oliver audit security groups for public access on ports 22, 3306, 5432

# Single agent — database
@tony analyze slow queries on production PostgreSQL

# Single agent — Kubernetes
@kai check pod resource utilization in the payments namespace

# Multi-agent — coordinated by Anna
@anna coordinate quarterly infrastructure review:
  - @alex: cost trends and savings opportunities
  - @oliver: security posture and compliance gaps
  - @tony: database performance health
  - @kai: Kubernetes efficiency

# Multi-agent — incident investigation
@anna parallel investigation:
  @alex check infrastructure and load balancers
  @tony analyze database performance
  @kai review pod health and networking
```

***

## How Agents Work

<Steps>
  <Step title="Receive your prompt">
    You mention an agent with `@agent` and describe what you need in natural language. Add `#tool` to specify the output format (dashboard, report, recommendations, alert).
  </Step>

  <Step title="Access your connections">
    The agent reads your connected infrastructure — AWS, GCP, Azure, Kubernetes clusters, databases — using the read-access credentials you've configured.
  </Step>

  <Step title="Analyze with context">
    Agents correlate data across multiple sources. [Alex](/guide/agents/alex) cross-references CloudWatch utilization with Cost Explorer data. [Oliver](/guide/agents/oliver) maps IAM policies against security group configurations. [Tony](/guide/agents/tony) reads execution plans alongside query frequency.
  </Step>

  <Step title="Generate output">
    Results are delivered in the format you requested — interactive dashboards, detailed reports, prioritized recommendations, or configured alerts.
  </Step>

  <Step title="Act with approval">
    If you need implementation, agents propose specific changes and wait for your [approval](/guide/approval) before executing. All actions are logged with an audit trail.
  </Step>
</Steps>

***

## Agent Autonomy Levels

You control how independently agents operate. Each level is configurable per agent:

| Level               | Mode                  | What Happens                                             |
| ------------------- | --------------------- | -------------------------------------------------------- |
| **L1 — Notify**     | Report only           | Agent surfaces findings, takes no action                 |
| **L2 — Suggest**    | Recommend             | Agent creates actionable recommendations for your review |
| **L3 — Approve**    | Act with confirmation | Agent proposes specific changes and waits for approval   |
| **L4 — Autonomous** | Self-directed         | Agent implements approved operations automatically       |

<Card title="Configure Agent Autonomy" icon="gear" href="/guide/approval">
  Set approval requirements per agent and per tool type
</Card>

***

## Memory and Context

Each agent maintains persistent memory of your environment:

* **Episodic memory**: remembers past analyses and decisions (e.g., which resources you've chosen to exempt from recommendations)
* **Working memory**: carries context within a conversation thread
* **Semantic memory**: stores patterns learned about your infrastructure over time
* **File memory**: retains documents, runbooks, and [knowledge base](/guide/knowledge) entries

This means agents get more accurate and useful over time as they learn your environment.

***

## Connection Requirements

Agents activate as you add the relevant connections:

| Agent      | Activates When                                        |
| ---------- | ----------------------------------------------------- |
| **Anna**   | Always available — no connections required            |
| **Alex**   | AWS, GCP, or Azure connection added                   |
| **Oliver** | AWS, GCP, or Azure connection added                   |
| **Tony**   | PostgreSQL, MySQL, or other database connection added |
| **Kai**    | Kubernetes cluster connection added                   |

<Card title="Set Up Connections" icon="plug" href="/guide/connections/overview">
  Connect your cloud providers, databases, and Kubernetes clusters
</Card>

***

## What's Next

<CardGroup cols={2}>
  <Card title="Alex — Cloud Engineer" icon="cloud" href="/guide/agents/alex">
    Cost optimization, infrastructure analysis, multi-cloud
  </Card>

  <Card title="Oliver — Security" icon="shield-check" href="/guide/agents/oliver">
    Compliance, vulnerability assessment, IAM, threat detection
  </Card>

  <Card title="Tony — Database" icon="database" href="/guide/agents/tony">
    Query optimization, performance tuning, analytics
  </Card>

  <Card title="Kai — Kubernetes" icon="dharmachakra" href="/guide/agents/kai">
    Cluster management, workload optimization, troubleshooting
  </Card>

  <Card title="Anna — General Manager" icon="users" href="/guide/agents/anna">
    Multi-agent coordination, executive reporting, complex ops
  </Card>

  <Card title="CloudThinker Language" icon="code" href="/guide/language">
    Complete `@agent #tool` syntax reference
  </Card>

  <Card title="Customize Agents" icon="wand-magic-sparkles" href="/guide/agents/customize">
    Tune personality, attach connections, and build fully custom agents
  </Card>
</CardGroup>