SStruct

Struct Turns Each Engineering Alert Into a Finished Investigation

Struct finished #2 of the day on Product Hunt on March 14, 2026 with 315 points, five days after its Y Combinator launch post. The product is an on-call agent: when an alert fires, it reads metrics, logs, traces, monitors and code, and returns a root cause, an impact summary and a draft incident report, often within 5 to 10 minutes by Struct’s figures.

Struct suits engineering teams whose on-call rotation spends hours tracing alerts by hand. The Startup tier is free for up to 5 users and 30 investigations a month, and setup takes minutes. It works best with structured logs and trace IDs. Struct scores 8.8/10 on our scale.

Struct’s Launch Results, Team and Claimed Metrics

Struct’s Y Combinator pages, its Product Hunt launch and its own product article supply these facts. Metrics marked as claims come from the company.

A #2 daily finish with 315 points is a good result for a three-person company. Product Hunt voters skew toward developers, the same people who carry pagers, so the result says the problem resonates with them.

The performance numbers are Struct’s own. A helpful rate of 85% to 90% would mean roughly one investigation in ten misses. Teams should measure that rate on their own alerts during the pilot, since it will depend on how clean their logs and traces are.

What Happens Between the Alert and the Root Cause

Struct sits between the alert source and the on-call engineer. Its job is the first hour of an incident, when someone usually opens five tools to work out what changed.

Connecting Alerts, Code and Telemetry

Setup links three things: an alert source (Slack or PagerDuty), GitHub, and one or more observability tools. Struct lists Datadog, AWS CloudWatch, Google Cloud, Azure and Sentry. The Y Combinator launch post says setup takes about five minutes, and Struct’s own article says under ten.

The Investigation Itself

As soon as an alert fires, Struct queries metrics, logs, traces and exceptions, and reads the recent commit history in GitHub. It correlates request IDs across cloud logs, runs regression analysis and lines up anomalies. Related alerts are grouped, so one database failure does not trigger five separate war rooms.

Alert Grouping and Noise

Struct’s launch post lists noisy alerts that never get tuned among the problems it targets. By grouping related alerts into one investigation, it keeps a single failure from opening several threads. Over time, the investigation history also shows which alerts fire often without a real cause, which gives the team a list of monitors to fix.

Runbooks From Senior Engineers

Teams can write their senior engineers’ checks as runbooks, and Struct runs them on every matching alert. A junior engineer on call gets the same first steps a staff engineer would take, without paging that staff engineer at 3 a.m.

Dashboards, Reports and Slack Follow-Up

Each investigation produces a dashboard with a unified timeline, the blast radius and the root-cause findings. Struct also drafts a full incident report with charts, timelines and commit histories. The summary and a link appear in the alert’s Slack thread, and a chat bot answers follow-up questions there.

Handing the Fix to a Coding Agent

Once the cause is confirmed, Struct can pass the context to a local CLI or a coding agent such as Claude Code, or generate a pull request itself. A person still reviews and merges the change.

A Checkout Error Spike, Step by Step

At 2:10 a.m., Datadog fires an alert for rising checkout errors. Struct starts at once: it pulls the error rate, finds matching exceptions in Sentry, follows the request IDs through CloudWatch logs and spots a deploy 20 minutes earlier that changed the payment client timeout. By 2:18, the Slack thread has the root cause, the share of users affected and the commit. The on-call engineer reads it on her phone, asks Struct to hand the fix to Claude Code and approves the revert pull request.

Struct Plans by Investigation Volume

Struct prices by the number of investigations a month. The Startup and Growth limits come from Struct’s own product article.

Matching Alert Volume to a Struct Tier

Thirty investigations a month is about one a day. A small startup with one or two real alerts a day fits the free tier if it routes only paging alerts to Struct.

A team that sees 7 alerts a day fills about 210 investigations in a 30-day month, slightly over Growth’s 200. Grouping helps here: if Struct folds related alerts into one investigation, the count drops. A team with 20 alerts a day, about 600 a month, is in Enterprise territory or needs to tune its alerting first.

The time math is the other side. If an engineer spends 30 minutes tracing each alert and Struct removes most of that, 200 investigations a month frees up close to 100 engineer-hours, using Struct’s claimed 80% reduction as the ceiling. Teams should replace that claim with their own numbers from the pilot. Struct could help buyers by publishing the Growth price, and by showing how grouped alerts count toward the monthly limit.

Struct Compared With Resolve AI, Oz and Kentik AI Advisor

All four help engineers respond to production problems. They differ in scope and in how they are sold.

Struct’s free tier and self-serve setup set it apart from Resolve AI, which sells by quote to larger organizations. Oz can run incident-response agents but asks the team to design them. Kentik AI Advisor focuses on networks. More options are in our IT and DevOps agents category.

On-Call Teams That Gain Most From Struct

Struct fits teams of 5 to 50 engineers where on-call rotates through people with uneven experience. Runbooks let the senior engineers’ first steps run on every alert, so juniors escalate less.

Teams where the same few senior engineers get pulled into every incident gain in a second way. Struct’s report and timeline give the next person on call the full context, so knowledge spreads beyond the people who wrote the original systems.

It also fits startups that want to delay a large observability purchase. Struct’s Y Combinator post names that use case directly. Teams already on Datadog, CloudWatch or Sentry with GitHub and Slack have everything Struct needs on day one. Healthcare software teams can use it under HIPAA.

Where Struct Fits in an Incident Tool Stack

Struct works on top of paging and monitoring and replaces neither. Datadog, CloudWatch or Sentry still detect the problem, and Slack or PagerDuty still wake someone up. Struct fills the gap between the page and the fix, the stretch where an engineer opens dashboards, searches logs and reads recent commits.

Teams that already use an incident management platform keep it for status pages, roles and postmortems. Struct’s draft incident report, with charts, timelines and commit histories, can feed that postmortem. The coding-agent handoff then links incident response to the repair, so the context gathered at 2 a.m. reaches whoever writes the fix.

Teams That May Want More History Before Buying

Large enterprises with formal vendor reviews may want more public customer results first. Struct has a three-person team and two named customers, and its performance numbers are its own.

Teams with unstructured logs and no trace IDs will get less precise answers, as Struct itself notes. Fixing logging first will make any AI investigation tool more useful. Teams that need on-premises deployment should confirm the sidecar and on-premises options with Struct, since they are listed for Enterprise.

A 30-Day Struct Pilot Scorecard

Struct offers a 30-day pilot with onboarding. A short scorecard makes the decision at the end clear.

Track four numbers for every alert Struct investigates. First, was the root cause right, partly right or wrong? Second, how many minutes passed between the alert and Struct’s summary? Third, how long did the on-call engineer spend after reading it? Fourth, did anyone senior get paged?

Compare those against the month before the pilot, using incident tickets or Slack history as the baseline. If Struct is right on most alerts and the escalations to senior engineers drop, the Growth tier has a clear case to a manager. If accuracy is low, check log structure and trace IDs before blaming the tool, since Struct says those drive precision.

Struct Data Handling and Security Answers

Struct is SOC 2 Type II and HIPAA compliant. It says logs and telemetry are processed ephemerally, are not stored and are not used to train models for other customers.

Since Struct reads production logs and code, a security review should still ask which scopes the GitHub app requests, how cloud credentials are stored and whether investigations can be limited to certain services. Read-only access to observability tools is enough for investigation; write access is needed only if the team wants Struct to open pull requests.

Struct Launch Week and Product Hunt Feedback

Struct has no star ratings on review sites yet. Its Product Hunt launch drew 315 points and 544 followers. One early user said it changed how their team handled the alerts channel. Commenters asked how Struct remembers patterns across incidents, how it handles outages caused outside the team’s own code and how it treats incidents spanning many services.

Written answers in Struct’s documentation would help buyers compare. Struct’s claimed 85% to 90% helpful rate is linked to Product Hunt feedback in its own article, which shows the team is collecting user ratings internally.

Those are fair questions for any investigation agent. Struct could answer them with public case studies from FERMAT or Arcana showing alerts handled, accuracy and time saved.

Struct’s 2026 Launch Sequence

• March 9, 2026: Struct launches on Y Combinator with the pitch to automate the on-call runbook.
• March 14, 2026: Struct finishes #2 of the day on Product Hunt with 315 points.
• 2026: Struct’s product article lists Startup, Growth and Enterprise tiers and SOC 2 and HIPAA compliance.

Struct Earns Its Score on Setup and Security

Struct scores 8.8/10. Setup and security lead: a free tier, minutes to connect and SOC 2 Type II plus HIPAA. Autonomy is high for investigation work. User sentiment sits at a neutral score because no star ratings exist yet, and pricing value is held back by the unpublished Growth price.

For a small or mid-size team tired of 3 a.m. log hunts, Struct costs nothing to test on the free tier. Route only paging alerts to it for a month, record how often its root cause was right and decide on Growth from those numbers.

On-Call Leads Ask These Questions About Struct

Is Struct Free?

The Startup tier is free for up to 5 users and 30 investigations a month, with no credit card. Growth and Enterprise cost more, with prices given on request.

Which Tools Does Struct Connect To?

Datadog, AWS CloudWatch, Google Cloud, Azure, Sentry, GitHub, Slack, PagerDuty and Linear, plus handoff to coding agents such as Claude Code.

How Fast Does Struct Find a Root Cause?

Struct says first-pass investigations take about 5 to 10 minutes. Speed and precision depend on how structured the team’s logs and traces are.

Does Struct Store Our Logs?

Struct says it processes logs and telemetry ephemerally, does not store them and does not use them to train models for other customers.

Does Struct Work With PagerDuty?

Yes. Struct lists PagerDuty and Slack as alert sources. An alert from either starts an investigation automatically.

Who Is Behind Struct?

Deepan Mehta and Nimesh Chakravarthi founded Struct, which went through Y Combinator’s Fall 2024 batch. The company is based in San Francisco.

Can Struct Fix the Problem Itself?

It can generate a pull request or hand the context to a coding agent. An engineer reviews and merges the fix.