IncidentOS

Problem Identified: We identified a major problem in modern e-commerce and software operations: when customers face issues while shopping, engineering teams often struggle to quickly identify the actual root cause and resolve it.

In a typical online shopping experience, customers may encounter problems such as products failing to load, incorrect availability information, add-to-cart errors, checkout failures, or unsuccessful payments. Customers usually see only a generic error message and have no clear way to report the problem with sufficient technical context.

At the same time, developers and Site Reliability Engineers (SREs) must investigate these incidents across multiple sources, including application logs, system metrics, deployment history, backend services, and source code. This investigation is often manual, time-consuming, and difficult to coordinate. A seemingly simple customer complaint, such as “My payment failed” or “The product is not loading,” may require engineers to examine several systems before identifying the real cause.

Who experiences this problem?

1. Customers

Customers experience:

• Products that are unavailable or fail to load.
• Missing product details or broken images.
• Add-to-cart and checkout failures.
• Payment errors.
• Lack of transparency after submitting a complaint.
• Delayed updates about whether their issue is being addressed.

Customers need a simple way to report problems directly from the shopping website without being redirected to an internal engineering dashboard.

2. Developers and SRE teams

Engineering teams experience:

• Incomplete customer reports.
• Difficulty reproducing customer issues.
• Time-consuming log and metric investigation.
• Challenges correlating customer complaints with recent deployments.
• Delays in identifying the faulty service or source-code line.
• Manual coordination between incident reporting, debugging, patching, and verification.

When multiple incidents occur simultaneously, these challenges can increase resolution time and potentially affect more customers.

3. Engineering and business teams

Organizations experience:

• Increased customer dissatisfaction.
• Revenue loss caused by checkout or payment failures.
• Longer incident resolution times.
• Repeated manual debugging work.
• Limited visibility into the relationship between customer-facing issues and technical failures.

How IncidentOS addresses this problem

IncidentOS connects the customer reporting experience of ShopOS with a separate internal engineering platform. When a customer reports an issue, the report is automatically registered as an incident in IncidentOS while the customer remains on ShopOS.

The engineering team can then use Claude to investigate the incident by analyzing simulated logs, metrics, deployment history, and source code. IncidentOS helps organize the investigation, identify possible root causes, highlight suspicious code, propose a patch, run verification tests, and generate an incident report.

The goal is to reduce the gap between “A customer is experiencing a problem” and “The engineering team understands, verifies, and resolves the underlying issue.”

Solution Overview

IncidentOS is an AI-powered incident investigation and remediation platform designed to connect customer-reported issues with engineering teams and help developers identify and resolve technical problems faster. It uses Claude’s reasoning and tool-calling capabilities to investigate incidents, analyze evidence, identify potential root causes, suggest code fixes, and verify those fixes through automated tests.

The platform consists of two separate experiences: ShopOS, a fictional customer-facing shopping website, and IncidentOS, an internal engineering dashboard.

On ShopOS, customers can browse products, view product availability, add products to their cart, and complete a simulated checkout. The website intentionally demonstrates realistic issues such as unavailable products, products that fail to load, missing product details, add-to-cart errors, and payment failures. When customers experience a problem, they can submit a report directly from the shopping website without being redirected to the engineering dashboard.

The report automatically captures relevant information, including the product name, issue category, product URL, transaction ID when applicable, customer description, timestamp, and other available context. After submission, the customer receives a confirmation and remains on ShopOS. Meanwhile, the issue is automatically registered in IncidentOS with a unique incident ID and a status such as OPEN.

The engineering team can access the separate IncidentOS dashboard to view and manage all incoming incidents. Each incident includes important details such as the reported problem, affected product, severity, customer impact, and issue category. This gives developers a structured and centralized view of problems instead of relying on incomplete customer messages or manually collected information.

The core functionality of IncidentOS is its Claude-powered investigation workflow. When a developer selects an incident and clicks “Investigate with Claude,” the AI agent investigates the issue using simulated engineering tools, including:

• Log analysis: Searches application logs for errors, timeouts, and failures.
• Metrics analysis: Reviews error rates, memory usage, latency, and service health.
• Kubernetes event analysis: Examines pod restarts, deployment events, and service warnings.
• Deployment history: Identifies recent releases that may have introduced the issue.
• Source-code investigation: Searches and reads the relevant files in the codebase.
• Hypothesis comparison: Evaluates multiple possible causes using available evidence.

Instead of immediately assuming one cause, Claude compares different hypotheses and explains why specific evidence supports or weakens each possibility. For example, if payment failures begin after a recent deployment and coincide with increased memory usage, Claude can connect these signals and identify a potentially problematic section of the payment service.

Once a probable root cause is identified, IncidentOS opens the relevant source-code file in an integrated code editor. The suspected faulty line is highlighted, and Claude proposes a possible fix. Developers can review a before-and-after code diff rather than accepting changes blindly.

The platform also includes a verification workflow. After reviewing the proposed patch, the developer must explicitly approve it before it is applied. The system then runs simulated unit, integration, and regression tests and displays the results. This provides a controlled process for validating the proposed solution before marking the incident as resolved.

After successful verification, IncidentOS generates a structured incident report containing the incident timeline, customer impact, investigation evidence, root cause, remediation details, and test results. A separate approval step can be used before sending a notification through a safe demo email workflow.

How IncidentOS Solves the Problem

IncidentOS addresses the gap between customer experience and engineering response. Customers receive a simple way to report issues while remaining on the shopping website. Developers receive structured, actionable incident information in a dedicated engineering dashboard.

Traditional incident investigation often requires engineers to switch between logs, monitoring systems, deployment records, and source-code repositories. IncidentOS brings these investigation steps into one workflow and uses Claude to connect evidence, explain possible causes, and recommend remediation.

The platform does not aim to replace developers. Instead, it acts as an AI investigation assistant that reduces repetitive investigation work, improves the quality of incident context, and helps engineers move from a customer complaint to a verified technical solution more efficiently.

The project demonstrates a complete incident lifecycle:

Customer Issue → Automated Report Registration → Claude Investigation → Root Cause Analysis → Code Patch → Test Verification → Incident Report

IncidentOS uses a realistic, controlled demo environment rather than real customer data, production infrastructure, or live payment systems. This makes it suitable for demonstrating how AI can support modern SRE and developer workflows in a safe and practical way.

It uses Claude’s reasoning and tool-calling capabilities to investigate incidents, analyze evidence, identify potential root causes, suggest code fixes, and verify those fixes through automated tests. The core functionality of IncidentOS is its Claude-powered investigation workflow. When a developer selects an incident and clicks “Investigate with Claude,” the AI agent investigates the issue using simulated engineering tools, including: Instead of immediately assuming one cause, Claude compares different hypotheses and explains why specific evidence supports or weakens each possibility. For example, if payment failures begin after a recent deployment and coincide with increased memory usage, Claude can connect these signals and identify a potentially problematic section of the payment service. The suspected faulty line is highlighted, and Claude proposes a possible fix.