AI debug engineer

Know why it broke.

An AI debug engineer finds why a customer got stuck and opens the pull request. You review it.

Card required. 3 days before you are billed.

Read-only. Opens a pull request. Never merges.

2h
back per small fix
1
pull request to review
0
merges without you
fernhill/checkoutWaiting for a failure
…

Nothing open yet.

Works with the stack you already run

ReactfrontendNext.jsfrontendVuefrontendNuxtfrontendSvelteKitfrontendAngularfrontendScript tagfrontendExpressbackendNestJSbackendFastifybackendHonobackendKoabackendFastAPIbackendDjangobackendFlaskbackendnet/httpbackendGinbackendEchobackend

One SDK for the browser, one for the server. Same visit.

01 /The problem

A customer said it broke.

The message does not say who, or which line. Finding that out takes hours. Then you still write the fix.

  1. 01

    Logs

    An error. No name on it.

  2. 02

    The page

    What they clicked is in another tool.

  3. 03

    GitHub

    The file you have open may not be the one that failed.

  4. 04

    The fix

    After all that, you still write it yourself.

Before and after

Without Bugwalk

With Bugwalk

How it reads

An AI debug engineer does those hours. You review the pull request.

02 /Same bug

An afternoon, or one review.

Drag to compare.

With Bugwalk

  1. 2:04 pmA message says checkout broke.
  2. 2:05 pmTheir visit is open. The failure is on it.
  3. 2:07 pmThe pull request is open against the live version.
  4. 2:15 pmYou review it and merge.

Without Bugwalk

  1. 2:04 pmA message says checkout broke.
  2. 2:20 pmYou ask what they clicked. They are not sure.
  3. 3:05 pmLogs, page and code do not match.
  4. 4:10 pmYou find the line and write the fix.
03 /What you get

What you get from one failure.

Their visit, the reason, and a pull request you review. The numbers are what a plan includes.

  1. 01

    2h

    Per small fix

    About two hours back, minus the time it took to open the request.

  2. 02

    100

    Investigations

    A month on Pro. 1,000 on Team.

  3. 03

    5

    Pull requests

    A month on Pro. 20 on Team.

  4. 04

    $39

    For about ten hours

    Team is $149, about forty hours. You still merge.

Bugwalk does not push to the default branch, comment, or merge.

04 /Try a failure

Pick a ticket. Watch it get fixed.

Three real-looking support messages from a small shop. Choose one, press the button, and watch the agent read the code.

Choose a ticket, then press the button.

fernhill/checkoutWaiting for a failure
…

Nothing open yet.

03 /The pull request

The fix is already written.

An AI debug engineer read what they did, named the line that was live, and opened the pull request.

  1. Read the visit

    What they did, and where it failed.

  2. Fix the code

    The line that failed, and nothing else.

  3. Open the pull request

    The request is ready before you debug.

  4. You review it

    You decide when it merges.

fernhill/checkout

Skip a coupon the discount table does not know

src/checkout/total.ts

- return subtotal * (1 - discount.percentage / 100);
+ if (!discount) return subtotal;
+ return subtotal * (1 - discount.percentage / 100);
  1. Read the visit
  2. Fix the code
  3. Open the pull request
  4. You review it

About two hours, looking it up by hand

5 places

Logs · Requests · What they did · GitHub · The live version

Reviewing the one request

Five places to check, then one request to review.

05 /One chain

From a name to a line of code.

Five links, kept together on one visit. Break any one and you are back in five tabs.

  1. 1

    Who

    The email they sent you opens their visits.

  2. 2

    What they did

    Every click, in order, on one timeline.

  3. 3

    What your app did

    The request and the error, on the same visit.

  4. 4

    Which version

    The release that was live, not the file on your screen.

  5. 5

    The fix

    A pull request written against that version.

05 /The visit

One place, not three tools.

What they clicked, what the app did, and the error stay on one visit.

[email protected] · checkout
  • 01

    The click and the request stay together.

  • 02

    The error stays on that same visit.

  • 03

    When they say it broke, you open their visit.

06 /The evidence

Check it. Do not just trust it.

Every reason points at something that happened, or at a line in your code. If the proof is thin, the report says so.

Finding
07 /Issues

The same bug, many people.

See who hit it, what they did, and why. Fix it once.

Issues
AllOpenRegressedResolvedIgnored
08 /The maths

Does it pay for itself?

Move the sliders. The only number we supply is two hours per small fix.

Pro opens 5 pull requests a month and Team opens 20. Past the plan, you keep the investigations and write the fix yourself.

Hours back a month

10h

Worth of time back

$800

Pro, per project costs $39. About $761 a month ahead after the plan.

09 /Live demo

A real app. Real failures.

Fernhill is a small shop we run. People sign in, check out and sometimes hit a bug. Open its dashboard and read a failure, no account needed.

  • People, sessions and issues, updating as the shop gets traffic.
  • Follow one shopper from the click to the error.
  • Break checkout yourself and see it land.
Open the live demo

Read-only. Investigations and pull requests are for your own project.

fernhill · live
  • AAvaPOST /checkout 500Open
  • BBenTypeError: address is nullOpen
  • CChengateway: token_expiredOpen
10 /Inside a project

Six places. One product.

01

Overview

The last day: who showed up, what failed, and which visits just ended.

02

People

Search the email or id you already have, then open someone and read their visits.

03

Session

One visit: the pages, what they did, what your app did, and where it failed.

04

Issues

The same failure, grouped: who hit it, what they did, and why.

05

Investigation

A written answer: what happened, why, how sure it is, and what to change.

06

Project

How Bugwalk is connected to your product, what it may read, and your plan.

11 /Setup

Add it to your product. Then watch the next failure.

Connect your app and the code it runs, then watch the next failure on the version that was live.

React, frontend:

$npx @bugwalk/wizard

Java, .NET, Ruby, and PHP

Point your existing OpenTelemetry agent at the Bugwalk OTLP endpoint. No Bugwalk library goes into those services at all.

$OTEL_EXPORTER_OTLP_ENDPOINT=https://api.bugwalk.app
12 /Privacy

It can look. You still merge.

It can read the failure and open a pull request. It cannot merge, or change your database.

Secrets stay in your app

Passwords, card numbers, and API keys are removed before anything is sent. You cannot turn that off.

A new branch only

It can open a pull request. It cannot push to your main branch, comment, or merge.

The SDK does not read

  • Keystrokes
  • Form values
  • localStorage
  • sessionStorage
  • IndexedDB
  • Environment variables
Before the batch leaves the app
13 /Questions

Before you connect a repository.

Can it change my code without me?

No. It can open a pull request on a new branch that starts with bugwalk/fix-. It cannot push to your default branch, comment, merge, or write to your database.

What does it read?

The visit that failed and the repositories you choose, at the commit that was live. Passwords, card numbers and keys are removed in your app before anything is sent.

What if it is not sure?

The report says so. A finding that cites nothing it actually saw drops the whole report to low confidence, with the reason shown.

How is it priced?

Per project. Pro is $39 a month, Team is $149. Usage stops at the plan limit and is never billed extra.

Can I run it myself?

Yes. The server and SDKs are Apache 2.0. The hosted product adds accounts, the agent and pull requests.

Open source, and hosted if you’d rather not run it. Read how to self-host.

Stop opening five tools.

Review the pull request.

Card required. 3 days before you are billed.

Pro is $39 · 3-day trial