Welcome to TraceX
TraceX is an AI engineer for production issues. It investigates across your databases, code, logs, errors and deploys, and tells you the root cause, the evidence and the fix.
What is TraceX?
Section titled “What is TraceX?”TraceX is an AI engineer for production issues. When something breaks, a customer writes in, an alert fires, or a teammate asks “why is this number wrong?”, TraceX investigates it the way your most experienced engineer would. It reads the code, runs read-only queries against your databases, searches your logs and errors around the time the problem started, and checks what was deployed just before. Then it writes up three things:
- The root cause: what is wrong, where (the table, the file and line, the service) and why it produces the symptom people see.
- The evidence: the queries and their results, the log lines, the errors, the code and the commit, so anyone can check the answer instead of trusting it.
- The fix: the code change, data correction or configuration change, how to check that it worked and, when you ask, a pull request.
Issues reach TraceX wherever your team already reports them: the dashboard, Slack, a support inbox, Jira, Linear, Zendesk, Intercom, PagerDuty, Opsgenie, your terminal or your AI coding tool. New to the words on these pages? Key concepts defines each one in a line.
Why use TraceX?
Section titled “Why use TraceX?”Most of the time spent on a production issue goes into finding the cause, not fixing it. Someone has to work out which system is involved, get access to it, write the right query, find the right log lines, read code they didn’t write and line it all up against what changed. That usually means pulling in the one engineer who knows the system, and interrupting whatever they were doing.
TraceX does that finding part. Some examples of what that looks like:
| An issue like | What TraceX does |
|---|---|
| “Checkout payments are failing since 09:02.” | Finds the failing requests in your logs, matches the error to the code that throws it, and ranks the commits deployed just before 09:02. |
“Order 48213 is stuck in pending_capture.” |
Looks the order up in your database, follows it through the payment code, and explains which step never ran and why. |
| “Signups dropped by half yesterday.” | Counts signups by hour and source in your database, compares them with errors and deploys, and points at the change that lines up. |
| “A customer says they were charged twice.” | Finds both charges, reads the retry logic that created them, and drafts a reply your support team can send. |
| “Is this pull request safe to merge?” | Reviews it with production context: your schema, recent incidents and your team’s rules, and comments on the lines that matter. |
Your team gets an answer with the proof behind it, so the job becomes checking an answer instead of hunting for one.
How is TraceX different?
Section titled “How is TraceX different?”Monitoring tools tell you that something is wrong: a graph turns red, an alert fires. Chat assistants answer from what they were trained on, or from documents you paste in. Neither connects the symptom to its cause in your systems.
TraceX goes into the systems an engineer would check during an incident and connects them:
- It reads your real data. A question about an order is answered from the order row, not from a guess.
- It reads your real code. Errors are matched to the file and line that throws them, and to the change that introduced them.
- It shows its work. Every claim cites evidence, and the report lists what TraceX ruled out, so you can see how far it looked.
- It is read-only and careful. Database connections are read-only, every query is checked before it runs, and secrets are masked before anything reaches an AI model. See How TraceX handles your data.
- It learns from your team. When someone corrects a root cause, TraceX remembers it for the next investigation.
Do I have to replace my monitoring or ticketing tools?
Section titled “Do I have to replace my monitoring or ticketing tools?”No. TraceX connects to the tools you already use, such as Datadog, Sentry, CloudWatch, Jira, Zendesk and PagerDuty, and posts its findings back into them. Your team keeps working where it already works.
Can TraceX change my database or my code?
Section titled “Can TraceX change my database or my code?”Not by itself. Database access is read-only and every query is checked before it runs. TraceX can write a fix and open a pull request, but a person reviews and merges it, and admins can require approval before any fix pull request is opened. See Fixes and pull requests.
What can it see?
Section titled “What can it see?”Only what you connect, per project, and only while it works on an investigation or a question. Every query, log search and file read is recorded in the data access log. You can disconnect everything at once with Settings → Revoke access.
How long does an investigation take?
Section titled “How long does an investigation take?”Most finish in a few minutes. You can watch it work live on the investigation page: the queries it runs, the searches it makes and what it finds.
What if TraceX is wrong?
Section titled “What if TraceX is wrong?”The report tells you how confident it is and what it ruled out, so a weak answer looks weak. When it is wrong, correct it at the bottom of the report: the correction becomes a fact TraceX uses in later investigations. See Read a report.
How do I get started?
Section titled “How do I get started?”Already have access? Sign in at app.tracexpro.com and follow the Quickstart: a project, one repository and one database is enough for a first investigation. New to TraceX? Request early access.