Skip to content

Talk through the bug. Get a report your team and your agent can use.

Voice2Bug is a Chrome extension. You click through the page and say what is wrong. It records your voice, your clicks, the pages you visit, console logs and network requests, and you take screenshots with one shortcut. When you stop, it builds a structured bug report. You can send it to Jira, Linear or email, share it as a link, or let a coding agent such as Claude Code or Codex read it through MCP.

THE WHOLE FLOW · WITH SOUNDA staged demo on a sample app (nothing is bought). Recording with voice and three screenshots, the report building itself, the expected result, sending to Jira and Linear, and Claude Code reading the report through MCP.
01 · Getting started

Getting started

What you need

  • Google Chrome 145 or later on macOS, Windows or Linux. Voice2Bug is a Chrome extension (Manifest V3). Other browsers are not supported.
  • A regular web page (http:// or https://). Voice2Bug cannot record on browser pages such as chrome:// settings, and it does not run in Incognito windows.
  • A microphone is optional. Without one, or if you deny access, the recording continues without audio and you can still take screenshots.
  • An activation code, which you get by email when you start a trial, or from your workspace administrator.

Install and activate Voice2Bug

ACCESS WINDOW · 18 SEnter your email, accept the consent, and send the confirmation email. Confirm the email in your inbox, then paste the activation code we send and click Activate Voice2Bug. The status changes to ACTIVE.
  1. Add Voice2Bug to Chrome from the Chrome Web Store and pin it to the toolbar so the icon is always visible.
  2. Open the Access window: click Start a trial on the floating Report a bug button or in the Side Panel. The window shows Current status with a badge: CHECKING, SETUP, ACTIVE, EXPIRED or ERROR.
  3. No code yet? Under Don't have an activation code?, enter your email, tick the consent checkbox and click Send confirmation email. Confirm the address from your inbox. If the address is eligible, we send your activation code.
  4. Have a code? Under Already have an activation code?, paste it and click Activate Voice2Bug. You will see Access activated on this device.

Good to know

  • The trial gives 30 days of full access with no card. Your 30 days start only after activation succeeds on this device, not when you request the email.
  • A trial is issued once per address: "No new license is issued after a trial has been used."
  • If your workspace asks for it, confirm your organization: in Confirm organization access, enter your work email, click Send code, then enter the code and click Confirm.
  • To sign this device out, use Manage this device → Remove access. Stop any active recording first.
  • Administrators can see the workspace activation code in the admin panel under Settings → Current activation code.
02 · Recording

Recording a bug

Start a recording

FLOATING BUTTON · 6 SClick Report a bug on the page. It shows Starting…, then turns red with a timer and STOP.

Open the page where the bug happens. You can start a recording in three ways:

  • the floating Report a bug button on the page;
  • the Voice2Bug icon in the Chrome toolbar (if this tab has no recording or report yet, one click starts recording);
  • the Report a bug button in the Side Panel.

Chrome asks for microphone access the first time. While recording, the button shows a red dot, the elapsed time and STOP.

Good to know

  • To move the floating button, open Settings → “Report a bug” button position and pick a corner. The choice is saved per site.
  • If your administrator limits recording to certain sites, other sites show: "This domain is not in your organization's allowed sites list."

Talk and click as usual

VOICE AND CLICKS · 5 SSay what you are doing while you use the page. Your clicks and page changes are recorded along with your voice.

Use the page the way you normally would and say out loud what you do, what you see, and what you expected. You do not have to type anything or switch windows. Your voice is transcribed and matched with your clicks, screenshots and logs by time.

  • Speaking is 3× faster than typing.
  • The spoken language is detected automatically. Report quality has been checked for English and Polish.
  • From version 3.9.41, a small voice meter next to the timer shows that your voice is being picked up.
  • If the microphone is unavailable, you will see "Recording without audio - microphone unavailable." and the recording continues.
  • The Side Panel can stay open while you record. Closing it does not stop the recording.

Take a screenshot

SHORTCUT · 4 SPress Option+Z on Mac (Alt+S on Windows and Linux), drag over the part you want, release. The counter shows Screenshot 1/10.
SystemWhile recordingWhen not recording
macOSOption+Z: select an areaOption+Z: open or close the Side Panel
Windows, LinuxAlt+S: select an areaAlt+S: open or close the Side Panel
  • Click and drag to select an area. Press Esc to cancel. If the area is too small, you will see "Selection too small".
  • Screenshots cover what is visible in the tab. There is no full-page capture.
  • You can take up to 10 screenshots per report. After that you will see "Screenshot limit reached (10/10). Stop the recording to build the report."
  • Each screenshot carries a small "Generated with Voice2Bug" mark.
  • To change the shortcut, open chrome://extensions/shortcuts.

Keep going across pages

TWO PAGES · 11 SThe recording follows you to the checkout page. Two more screenshots there, including the error banner that appears after Pay.

The recording continues when you move to another page in the same tab. Every page you visit is added to the navigation path, and you can take screenshots on any of them.

What is captured automatically

EvidenceWhat it containsKept in the report
Your voiceTranscript, with timing for each partReport transcription and original transcription
User actionsClicks, typing, key presses, scrollingActivity Logs → User Actions
Consolelog, info, warn, error and debug messagesUp to 80 entries; long messages are cut at 1,500 characters
Networkfetch and XHR requests with status and timingUp to 80 requests. Response bodies are kept for failed (4xx/5xx), slow (3 s or more) or data-changing requests, up to 64 KB.
NavigationEvery page address, in orderNavigation path
EnvironmentOS, architecture, viewport, browser (user agent), languageOS & Architecture, Recorder's Browser

Sensitive data and redaction

  • Typed values in password, email, phone, number, search and URL fields are not stored. They appear as "Typed redacted … value". The same applies to fields whose name suggests a token, password, PIN, card, CVV, IBAN, session, cookie or similar.
  • Request headers: only a short list of safe headers is kept. authorization, cookie and x-api-key are replaced with [REDACTED].
  • Request and response bodies: values that look like secrets are replaced before the report is saved.
  • Screenshots are not blurred automatically. Select only the area you need.
  • Administrators can limit recording to allowed sites and skip domains in network activity.

Where data is processed and stored is described in the privacy policy.

Stop and get the report

STOP · 7 SClick STOP. The button shows Building report, the Side Panel shows Building your report…, then Report ready.

Click STOP on the floating button or Stop and build report in the Side Panel. Voice2Bug collects the evidence, filters the logs and network errors, asks AI to draft the report and saves it. You can keep browsing while it works. If Chrome suspends the extension, the progress is restored.

If something goes wrong

  • Report needs attention: click Retry report, or Start new recording.
  • "This recording has too many attachments": record a shorter session or take fewer screenshots.
03 · The report

Working with the report

What you said, next to each screenshot

TIMELINE · 10 SDetailed Timeline / Observations: each observation with its Source timing and the screenshot taken at that moment.

The Detailed Timeline / Observations section turns your recording into short observations. Each one shows where it came from:

  • Source timing: m:ss - m:ss: the exact part of the recording;
  • Approximate timing or Timing unavailable when the match is less certain.

If the AI is not confident about part of the report, the report says so, for example "AI confidence is low. Check the report against the source."

Edit a step

STEPS TO REPRODUCE · 5 SClick a step, change the text, confirm with ✓.

The title, Steps to Reproduce and timeline items are editable. Click the text to change it. Edited items get an Edited badge. Removed items can be brought back: "Report item removed. Restore is available below." Use Revert in the local changes panel to undo your edits.

NOTEAfter manual edits the report shows "This report contains manual edits. Source links may no longer apply."

Add the expected result

EXPECTED RESULT · 6 SType what should happen. The counter shows the length, and the field saves itself: Saved.

Write what should have happened in the Expected result field (hint: "What should happen?"). It holds up to 1,200 characters and saves automatically. The status shows Saved, Unsaved changes or Save failed. The expected result is included when you send the report and when an agent reads it.

What is in a report

SectionWhat it shows
TitleOne line describing the bug. Click to edit.
Navigation pathPages visited, in order
OS & Architecture, Recorder's BrowserThe environment where the bug was recorded
Expected resultWhat should happen (written by you)
Steps to ReproduceNumbered steps (editable)
Detailed Timeline / ObservationsObservations with source timing and screenshots
Report transcriptionWhat you said, cleaned up, plus the original transcription
Activity LogsUser Actions and Console Logs
Network EvidenceRequests with filters: Shown, Failures, Slow, 4xx, 5xx. You can remove a request from the report and restore it.

Saved reports

Open the Side Panel → Saved reports to see reports kept in this browser. You can attach a report to the current tab, detach it, or remove it. Up to 50 reports (250 MiB) are stored locally. Voice2Bug warns you when you are close to the limit. Removing a report from the extension does not delete copies already sent to the server, Jira or email.

04 · Sending

Sending the report

All the ways to send

ONE REPORT · 7 SCopy All, Share with link, Send by email, Jira and Linear, all from the same report.
OptionWhat happens
Copy AllCopies the full report, with images, ready to paste into a ticket, chat or document. Toast: "Full report copied!"
Share with linkCreates a link to a report page. See Share with a link.
Send by emailSends to the inbox your administrator set up
JiraCreates an issue in the project and issue type you choose
LinearCreates an issue in the team set by your administrator
Coding agent (MCP)Claude Code or Codex reads the report directly. See Coding agents.

Send to Jira

JIRA · 13 SPick a project and an issue type, click Send to Jira. The issue arrives with steps, timeline, logs and screenshots.
  1. In the Jira card, choose Project and Issue type.
  2. Click Send to Jira. You will see "Created Jira issue KEY."
  3. After you change the report, the button becomes Update Jira ticket.

The issue contains the navigation path, environment, expected result, steps to reproduce, timeline, activity logs and network activity, with screenshots attached.

  • YOLO mode sends the report automatically after it is generated.
  • If you see "Delivery is unconfirmed. Check Jira before trying again.", look in Jira before you resend, to avoid a duplicate.
  • The Jira connection is set up by an administrator: Delivery settings.

Send to Linear

LINEAR · 11 SExpand the Linear card, check the team set by your admin, click Send to Linear.

The team is configured by your administrator, and the project is optional. Sending to Linear is always manual. Click Send to Linear, and later Update Linear ticket after changes. The issue has the same sections as in Jira, the screenshots, and a link back to the Voice2Bug report.

Send by email

EMAIL · 10 SThe destination is shown under the buttons. Click Send by email: "Email sent." The report arrives in the inbox with screenshots.

The destination is set by your administrator and shown under the buttons, for example Destination: QA inbox. If nothing is set, you will see Destination: not configured. Tick Send a copy to me to receive a copy. After sending, the button becomes Send again. If you see "Email delivery was not confirmed", check the inbox before you resend.

A helpdesk that accepts email (an intake address) can receive reports this way too.

05 · Coding agents

Coding agents and MCP

What your coding agent can read

MCP · 8 SThe report goes through the Voice2Bug MCP server (read-only, 8 tools) to Claude Code or Codex.

A coding agent cannot see your browser. Voice2Bug gives it the evidence through an MCP server: the summary, steps, console, network, user actions, navigation, transcript and screenshots. The agent starts when you prompt it, for example "Fix the latest Voice2Bug report."

The connection is read-only: "The connection cannot create, modify, or delete reports."

ToolReturns
v2b_list_reportsRecent reports in your workspace (1 to 20 per page)
v2b_get_report_summaryTitle, environment, expected result, steps, transcript excerpt, evidence counts and suggested next calls
v2b_get_report_evidenceOne category at a time: steps, console, network, actions, navigation, transcription, screenshots, timeline
v2b_get_console_logsConsole entries (redacted), paginated
v2b_get_network_requestsRequests and responses (redacted)
v2b_get_user_eventsUser actions or navigation
v2b_get_transcriptThe report transcript or the original source transcript
v2b_get_screenshotScreenshot list, or a single image

Connect Claude Code

SETUP · 12 SPaste the command from the admin panel, then sign in and approve read-only access in the browser. Once per developer.
  1. In the admin panel, open Connect coding agent → Claude Code and click Copy Command. The command already contains your workspace's profile name and server address.
  2. Paste it in your terminal. It has this shape:
    claude mcp add --transport http --scope user <profile> "<server URL>" && claude mcp login <profile>
  3. A browser page opens: Connect a coding agent to Voice2Bug. Sign in with your Voice2Bug account and approve.

Claude Code must be signed in first (claude auth login).

Connect Codex

Same steps, using the Codex tab in Connect coding agent:

codex mcp add <profile> --url "<server URL>" && codex mcp login <profile>

See Codex fixes the latest report.

OTHER MCP CLIENTSThe admin panel provides commands for Claude Code and Codex, the two clients we have tested.

Claude Code fixes the latest report

REAL SESSION · SPED UP · 34 S"Fix the latest Voice2Bug report." Claude Code reads the report through MCP, finds the cause, reproduces it, changes the code and runs the tests. The column on the right explains each step.

Example prompts:

  • Fix the latest Voice2Bug report.
  • Read Voice2Bug report <link or ID> and find the root cause. Don't change code yet.
  • Show me the failing network requests from the latest Voice2Bug report.

The agent decides which tools to call. It usually starts with the summary and then reads the console, network and screenshots it needs. Review its changes the same way you review any pull request.

Codex fixes the latest report

REAL SESSION · SPED UP · 27 S"Fix the latest Voice2Bug report." Codex reads the summary, console, network and screenshots through MCP, fixes the code and runs the tests.

The prompts are the same as for Claude Code.

Limits and disconnecting

  • An agent can read reports from the workspace it is connected to.
  • Limits: 60 requests per minute per connection. Screenshots up to 2 MB each.
  • To disconnect, open the admin panel → Connect coding agent → Active connections → Disconnect. Locally, run claude mcp remove --scope user <profile> or codex mcp remove <profile>.
06 · Admin panel

For administrators

Sign in and invite your team

Sign in to the admin panel with your email and password, or with a Magic Link. The Start page walks you through setup: confirm the organization, install the extension, activate it, create and check the first report, set up delivery, and send the first report.

Open Team → Invite User and enter an email, name and role:

RoleCan
UserRecord, generate, edit and send reports
AdministratorAlso manage the organization, users and integrations

The panel is in English by default. You can switch to Polish in Settings.

Delivery: email, Jira and Linear

In Integrations (Configure report delivery), click Add Channel:

  • Email: a name, the recipients or a helpdesk intake address, whether it is the default destination, the report sections to include, and automatic or manual delivery.
  • Jira Cloud: Connect to Jira, then choose the default project and issue type.
  • Linear: Connect to Linear, then choose the team and, optionally, a project.

Allowed sites

Choose where your team can record: All pages or Limited to the sites you list. On other sites, recording does not start and the user sees why. You can also list domains to skip in network activity, for example analytics or third-party services.

Trial and access

During the 30-day trial your workspace has full access and unlimited reports, with no card and no automatic payment. When the trial ends, recording, report content, exports and share links are blocked, but your account and data are not deleted automatically. The Access window and Organisation → Current plan show your status. To continue, contact support@voice2bug.com.

07 · FAQ

Questions and limits

Which browsers are supported?

Google Chrome 145 or later on desktop. Other browsers are not supported, and the extension does not run in Incognito windows.

Why can't I record on this page?

  • The page is not a regular web page (for example a chrome:// settings page). Open a normal http(s) page.
  • The tab was open before you installed the extension: "Refresh this tab or open a regular webpage, then click Voice2Bug again."
  • Your organization allows recording only on certain sites.
  • Local storage is full: "Local report storage is full. Remove a saved report before recording again."

Do I have to speak?

No. Voice and screenshots are optional. Without voice, the report is built from your actions, screenshots, logs and network requests. Speaking gives the report the "why": what you expected and what looked wrong.

Which languages can I speak?

The spoken language is detected automatically. Report quality has been checked for English and Polish.

Can the coding agent change or delete my reports?

No. The MCP connection is read-only. The agent can read reports, but it cannot create, change or delete them.

Where is my data stored, and for how long?

Captured data stays in your browser until you generate a report. Reports are kept until you or your administrator delete them, and share links work until they are revoked. Processors and storage locations are listed in the privacy policy.

No topics match your search. Try "screenshot", "Jira" or "MCP".