> ## 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.
# CloudThinker
> AI-powered multi-agent platform for autonomous cloud operations — cost optimization, code review, and incident response across AWS, Azure, and GCP
AI agents that manage your infrastructure, review code, resolve incidents, and optimize costs — across multi-clouds, Kubernetes, and everything in between. Self-healing infra, autonomous.
Connect your cloud and run your first operation in 5 minutes
Master the `@agent` and `#tool` syntax for effective prompting
***
## Why Cloud Operations Is Broken
Cloud teams juggle 8–12 separate platforms for cost, security, monitoring, and operations. When a service goes down at 3 AM, engineers spend more time correlating dashboards than fixing the problem. Cost reviews happen monthly. Security audits take weeks. And every new cloud service adds another point tool to manage.
**Three compounding problems:**
| Problem | Reality |
| ------------------ | ------------------------------------------------------------------------------------------------------------------------------------------ |
| **Too many tools** | Cost Explorer for spend, Security Hub for compliance, Datadog for monitoring, kubectl for containers — and none of them talk to each other |
| **Too slow** | Manual root-cause analysis takes hours; incidents stay open while engineers piece together the story across six dashboards |
| **Too hard** | Deep cloud expertise required just to ask basic questions — a permanent blocker for developers and non-specialists |
***
## The CloudThinker Difference
Most cloud management tools are dashboards — they show you data, but you still have to interpret it, decide what to do, and manually execute changes.
CloudThinker acts.
| Capability | Traditional Tools | CloudThinker |
| --------------------- | ----------------------------------- | ---------------------------------------------------------------------------------------------------- |
| **Cost visibility** | Show historical spend charts | Identify waste, recommend right-sizing, implement with approval |
| **Security posture** | Report misconfiguration counts | Investigate context, prioritize risk, generate remediation plans |
| **Incident response** | Page you and show metric graphs | Suppress 98% of noise upstream, then investigate root cause autonomously before you open your laptop |
| **Compliance** | Generate point-in-time scores | Produce evidence, close gaps, maintain continuous compliance |
| **Interface** | Console + CLI + separate dashboards | Natural language across all domains, any cloud |
Five specialized AI agents replace the fragmented tool stack — and they coordinate with each other on complex problems.
***
## Core Modules
AI-powered PR reviews with security analysis — 96% accuracy
Pulse cuts the noise, AI investigates the rest — MTTR under 5 minutes
Tiered support with intelligent escalation — 70% auto-resolved
Unified cost, security, and performance — 325+ operations
24/7 autonomous monitoring and compliance
Conversational ops directly in Slack
AI agents review every pull request for bugs, security vulnerabilities, and best-practice violations — then post actionable feedback directly in your PR. Security analysis runs automatically alongside code quality checks, catching issues before they reach production.
From cloud event to resolution, end to end. **Pulse** ingests 10+ sources and cuts \~98% of noise through seven suppression layers and AI classification. **Incident** investigates the moment a cluster escalates — forming hypotheses, testing them against evidence, and executing remediation runbooks under your approval gates. MTTR as low as **4m 32s**.
A multi-tier support system where AI agents handle routine requests while intelligently escalating complex issues to the right specialist. **70% of tickets are auto-resolved** without human intervention.
Three pillars in one unified interface: **Cost** optimization to eliminate waste, **Security** posture management to stay compliant, and **Performance** monitoring to keep systems healthy. Agents continuously analyze and recommend improvements across all three.
Always-on autonomous agents that continuously scan your cloud environment for drift, compliance violations, and optimization opportunities. [CloudKeepers](/guide/infrastructure/cloudkeepers) work around the clock — surfacing issues before they become incidents.
Run cloud operations directly from Slack with native conversational UI ([SlackOps](/guide/slack-integration)). Mention an agent, describe what you need, and get results — all without leaving your team's workspace.
***
## Platform Capabilities
The foundational layer that powers every module — composable, secure, and observable.
Composable skill definitions (SKILL.md, tools, prompts, guardrails, triggers)
4-level autonomy: Notify, Suggest, Approve, Autonomous — RBAC-gated
Ephemeral microVM environments with per-tenant VPC isolation
325+ pre-built operations with cron scheduling and chaining
MCP-based integrations — AWS, Azure, GCP, K8s, Slack, GitHub, Datadog, and more
Vectorized RAG knowledge base with continuous learning
Real-time resource mapping and dependency graphs across regions
Multi-layer memory: episodic, working, semantic, and file storage
PII detection, schema validation, and prompt injection defense
OpenTelemetry tracing, LLM-as-Judge evaluation, dashboards
Every agent capability is defined as a composable **Skill** — a combination of SKILL.md definitions, tool bindings, prompts, guardrails, triggers, and schemas. Skills are the building blocks that make agents specialized.
Control how independently agents operate: **L1 Notify** (report only), **L2 Suggest** (recommend actions), **L3 Approve** (act with approval), **L4 Autonomous** (full self-directed execution). Each level is gated by RBAC policies.
Agents execute in **ephemeral, isolated sandbox environments** — microVMs with per-tenant VPC isolation that auto-destroy after use. Agent actions never touch your production infrastructure directly.
Pre-built operational procedures agents execute step-by-step. **325+ operations** available, schedulable via cron, chainable for complex workflows, and customizable for your environment.
An **[MCP](/guide/connections/mcp)-based integrations hub** connecting agents to your infrastructure. AWS, Azure, GCP, Kubernetes, Slack, GitHub, Datadog, [Grafana](/guide/connections/grafana), and more — all through a standardized protocol.
Agents build and query a **vectorized [knowledge base](/guide/knowledge)** from your docs, runbooks, past incidents, and operational history. Continuous learning through RAG means increasingly accurate recommendations over time.
A **visual resource graph** mapping relationships across regions and providers. [Topology](/guide/infrastructure/topology) powers [Root Cause Analysis](/guide/incident/root-cause-analysis) — when something breaks, agents trace the impact path instantly.
**Multi-layer memory** gives agents persistent context: **Episodic** (past interactions), **Working** (current task), **Semantic** (learned concepts), and **File** (document storage). Agents remember your preferences and past decisions.
**Input and output safety gates** protect every agent interaction — PII detection, schema validation, prompt injection defense, and content filtering. Agents always operate within defined boundaries.
Full **platform monitoring and evaluation** powered by OpenTelemetry. Track agent performance, trace execution, evaluate output quality with LLM-as-Judge, and monitor system health through built-in dashboards.
***
## The Agent Team
Five specialized AI agents — always on, scalable, secure, intelligent.
| Agent | Role | Primary Focus |
| ---------------------------------- | ------------------------ | ----------------------------------------------------------- |
| **[Alex](/guide/agents/alex)** | Cloud Engineer | Cost optimization, infrastructure, multi-cloud architecture |
| **[Oliver](/guide/agents/oliver)** | Security Engineer | Compliance, vulnerabilities, threat detection, IAM |
| **[Tony](/guide/agents/tony)** | Database Administrator | Query optimization, performance tuning, data analytics |
| **[Kai](/guide/agents/kai)** | Kubernetes Administrator | Container orchestration, cluster optimization, workloads |
| **[Anna](/guide/agents/anna)** | Technology Leader | Multi-agent coordination, strategy, executive reporting |
```bash theme={null}
@alex analyze EC2 instances with <20% CPU utilization over 30 days
@oliver audit security groups for public access on database ports
@tony #dashboard database performance metrics for production cluster
@kai optimize pod resource allocation across all namespaces
@anna coordinate quarterly infrastructure review with all agents
```
***
## How It Works
Link your cloud accounts, Kubernetes clusters, and tools. Agents discover your infrastructure automatically.
Use natural language with `@agent` mentions and `#tool` commands to request analysis or actions.
Agents analyze, recommend, and — with your approval — implement optimizations autonomously. Or set them to full autonomy and let the system self-heal.
***
## What You Can Do
```bash theme={null}
@alex analyze spending trends over last quarter
@alex #recommend reserved instance purchases for stable workloads
@alex identify unattached volumes and unused elastic IPs
```
**Typical outcome**: 30–50% cost reduction with automated implementation plans
```bash theme={null}
@oliver perform SOC 2 Type II compliance assessment
@oliver audit IAM policies for privilege escalation risks
@oliver #report security posture with remediation timeline
```
**Typical outcome**: 90%+ compliance score with complete audit trail
```bash theme={null}
@tony analyze slow queries on production PostgreSQL
@tony #dashboard query performance with P95 latency trends
@tony recommend index optimizations for high-frequency queries
```
**Typical outcome**: 70% query performance improvement
```bash theme={null}
@kai analyze pod resource utilization across all clusters
@kai identify nodes with <30% CPU utilization for consolidation
@kai #recommend HPA policies for variable workloads
```
**Typical outcome**: 25% cost reduction with improved reliability
```bash theme={null}
@anna coordinate AWS to Azure migration with all agents
@anna #report quarterly infrastructure review for executive team
@anna manage security remediation project across @oliver @alex @kai
```
**Typical outcome**: 75% faster project completion with unified visibility
***
## Next Steps
Get running in 5 minutes with step-by-step setup
Set up integrations with your cloud providers and tools
Learn the complete syntax for agent and tool commands
See real-world examples of CloudThinker in action
