> ## 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.

# Deep Response Engine

> Learn how CloudThinker turns raw cloud signals into investigated, remediated incidents through the Pulse and Incident stages.

The Deep Response Engine is CloudThinker's incident lifecycle module. It carries every event from first signal to resolved incident — noise reduction, escalation, root cause analysis, remediation, and memory.

Most monitoring stacks tell you something is wrong and stop there. The Deep Response Engine tells you why: [Pulse](/guide/pulse/overview) filters and correlates signals before they page anyone, and Incident starts investigating the moment a cluster escalates — often before an on-call engineer opens a laptop.

## How it works

<div style={{display: 'flex', justifyContent: 'center'}}>
  ```mermaid theme={null}
  flowchart TD
      A["☁️ Cloud events\nAWS · Datadog · PagerDuty and more"] --> B

      subgraph B["PULSE"]
          direction TB
          B1["Noise reduction · correlation"]
          B2["AI classification · severity ranking"]
          B1 --> B2
      end

      B --> |escalate| C

      subgraph C["INCIDENT"]
          direction TB
          C1["AI investigation · root cause analysis"]
          C2["Automated remediation · memory"]
          C1 --> C2
      end
  ```
</div>

No stage requires a manual handoff — each layer feeds the next automatically:

1. **Collect** — events stream in from sources like AWS, Datadog, Slack, and PagerDuty into one Pulse feed.
2. **Filter and correlate** — suppression layers remove duplicates, rate-limited bursts, and flapping resources. Related signals are grouped into clusters, so nine alerts about the same node pool become one item.
3. **Classify and escalate** — every signal gets a category, canonical severity, and actionability score. When a cluster is Critical or High, or the AI marks it actionable, it escalates to an incident automatically.
4. **Investigate** — an AI agent forms explicit hypotheses, tests each one against metrics and logs, and produces a structured report: most likely root cause, evidence chain, and ruled-out theories.
5. **Resolve and remember** — the agent matches your [runbooks](/guide/incident/runbooks) to the root cause and executes them under the autonomy mode you set (Manual or Auto). Each resolution feeds [incident memory](/guide/incident/incident-memory), making the next investigation faster.

Every investigation step is visible — which hypothesis was confirmed, which was ruled out, and why.

## What you can do

| Capability                 | Description                                                        | Guide                                                                 |
| -------------------------- | ------------------------------------------------------------------ | --------------------------------------------------------------------- |
| Connect signal sources     | Feed AWS, Slack, Teams, and webhook events into Pulse              | [Pulse setup](/guide/pulse/setup)                                     |
| Manage signal clusters     | Review, merge, and act on correlated signal groups                 | [Clusters](/guide/pulse/clusters)                                     |
| Run AI root cause analysis | Follow hypothesis-driven investigations to a structured RCA report | [How it works](/guide/incident/root-cause-analysis)                   |
| Ingest monitoring webhooks | Route alerts from PagerDuty, Datadog, CloudWatch, and more         | [Webhook integrations](/guide/incident/webhook-integrations/overview) |
| Automate remediation       | Let agents execute matching runbook procedures                     | [Runbooks](/guide/incident/runbooks)                                  |
| Log incidents manually     | Record incidents that started outside Pulse                        | [Manual logging](/guide/incident/manual-logging)                      |
| Learn from every incident  | Reuse what worked — queries, techniques, runbook steps             | [Incident memory](/guide/incident/incident-memory)                    |
| Measure the loop           | Track noise reduction, cluster MTTR, and conversion rates          | [Pulse analytics](/guide/pulse/analytics)                             |

## Key concepts

| Concept         | Meaning                                                                   |
| --------------- | ------------------------------------------------------------------------- |
| Signal          | A single normalized event from any connected source                       |
| Cluster         | A group of correlated signals treated as one item                         |
| Incident        | The investigation object created when a cluster escalates                 |
| Runbook         | An operational procedure agents can match and execute during remediation  |
| Incident memory | The record of techniques, queries, and steps that resolved past incidents |

## Get started

<CardGroup cols={2}>
  <Card title="Connect signal sources" icon="satellite-dish" href="/guide/pulse/setup">
    Connect AWS, Slack, Teams, and webhook sources to start feeding Pulse.
  </Card>

  <Card title="Set up webhook integrations" icon="webhook" href="/guide/incident/webhook-integrations/overview">
    Route alerts from PagerDuty, Datadog, CloudWatch, and more into the response loop.
  </Card>

  <Card title="Add runbooks" icon="book" href="/guide/incident/runbooks">
    Give agents the procedures they can execute during remediation.
  </Card>

  <Card title="See how investigation works" icon="flask" href="/guide/incident/root-cause-analysis">
    Follow a hypothesis-driven root cause analysis end to end.
  </Card>
</CardGroup>
