Skip to content
FOR WEB APPS IN ALPHA AND BETA

Testers report. Your agent fixes.

Paste one snippet and invite your testers. Testosaurus groups what they report, including the frustrations that never throw an error, into ranked issues with page, console and consented screenshots. Your coding agent picks the top one over MCP, fixes it in your repo and runs your tests.

Building with a coding agent? Paste this into it:

Add Testosaurus to this app by following https://testosaurus.dev/docs/connect.md (read it with curl -fsSL).

It opens a page where you sign up or sign in, check the code and click Allow. Nothing secret to copy.

Free for one app: 100 reports and AI triage a month, no card.

Testo, your friendly green dinosaur, at the keyboard
Example reportRPT–024
BugCheckout New

Checkout button does nothing

“I click Pay and nothing happens.”

The message. And the evidence behind it.
acme✳Checkout
ALMOST YOURS

Complete your order

Everyday teeForest / Medium
€48.00
Subtotal
€48.00
Shipping
€5.00
Total
€53.00
Pay
Screenshot attached
Console 1 error
TypeError:
Failed to fetch
AI hypothesis

Check the failed checkout request.

A starting point to investigate.
Captured context/checkout 1440 × 900
FROM TESTER REPORT TO A FIX YOU CAN REVIEW.
ReportRankFix
01 / THE AGENT LOOP

From report
to a fix on a branch.

Bring your own coding agent. Testosaurus hands it the ranked issue and the evidence. The fix happens in your repo, with your tests, under your review.

  1. 01

    Testers report.

    They write what went wrong in your alpha or beta, including the frustrations that never throw an error. Page and console come along. Screenshots and replay are their choice.

  2. 02

    Testosaurus groups and ranks.

    Reports about the same problem become one issue, ranked by impact: severity first, then how many reports.

  3. 03

    Your agent fixes it.

    It runs the fix_top_issue prompt (Claude Code: /mcp__testosaurus__fix_top_issue, or /testosaurus:fix with the Claude Code plugin), shows you the evidence, fixes on a branch, runs your tests and records the fix back.

Connect your coding agent
Your agent.
Your review.

Your agent works the tester inbox.

Connect Claude Code, Codex or any MCP client. It reads the ranked issues with their evidence, asks before it edits, and writes the result back. With your own agent, nothing is pushed or merged without you.

Example: your agent’s sessionMCP

/mcp__testosaurus__fix_top_issue

list_issuesresult
Input

status: new · sort: impact · dir: desc

1 · Bug · High · 3 reports

Checkout button does nothing

“I click Pay and nothing happens.”

RPT–024
2 · Confusing · Medium · 2 reports

Promo code field gives no feedback

RPT–019

Ranked by impact: severity first, then how many reports. Counts are reports, not distinct people.

get_issueresult
Original feedback · 3 reports

“I click Pay and nothing happens.”

Captured context

ConsoleTypeError: Failed to fetch

AI hypothesis

Check the failed checkout request.

Available attachments

Tester text arrives fenced as untrusted data. Capture depends on consent, and AI interpretation stays separate from captured facts.

record_fixresult
Summary

Show a retry message when the checkout request fails.

Stateproposed
Tests runnpm test -- checkout · passed

Your agent asked before editing, worked on a branch and ran your own tests. Nothing was pushed or merged without you.

Evidence ready for your agent to work from.

Run it for me

Beta

Prefer not to drive an agent yourself? A coding role in your own GitHub Actions proposes a draft PR that has passed your configured tests. It never merges. It needs:

  • a Node.js 22 project
  • GitHub repo and workflow authorization
  • prepaid AI credit (the monthly included credit does not fund runs)

You start each run, or opt in to automatic runs for new reports, which open a draft PR without your review.

ExampleDraft PR
RPT–024 → Checkout fixes

Show a retry message when checkout fails.

Allowed files
src/checkout
Configured tests
npm test -- checkout
Run cap
€2 in this example
Configured tests passedDraft ready for human review

Example output, not a live run. Review the diff and test results before merging.

Set up managed runs
02 / IT STARTS IN YOUR APP

Feedback, right
where it happens.

An in-app feedback widget gives users a place to tell you what went wrong. Browser context comes along. Screenshots and replay are their choice.

  1. 01

    A familiar place to speak up.

    Choose Testo’s friendly peek, a classic bubble, or a simple feedback pill.

  2. 02

    Let your users choose.

    Recording, screenshot only, or neither. Useful feedback starts with clear consent.

  3. 03

    A few words. Useful context.

    One small message becomes a report your team can investigate.

See the live widget guide

AN EXAMPLE INSIDE YOUR APPInteractive preview
store.example/checkout
Send feedbackWe’re listening.
bugconfusingidea

Screenshot and replay are optional.

Powered by Testosaurus

Launcher, surface, placement and consent style are free. Accent colors, wording, tags and unbranded widgets are available on Pro. Explore customization

03 / GROUPED AND RANKED

A few words in.
An actionable report out.

Reports about the same problem share one issue. Your chosen AI model turns the feedback and captured evidence into a summary, reproduction steps and a hypothesis to investigate.

01

See what they saw.

The original message and its available evidence, together in one report.

02

Let AI find the signal.

The analysis makes each issue easier to act on, and the ranking puts the most damaging one first. Original feedback, captured facts and AI interpretation stay distinct.

03

Put the report to work.

Hand the issue to your coding agent over MCP, or create a GitHub or Linear issue with the context attached.

Preview the handoff
From the user’s words
to your team’s next move.
Example reportRPT–024
BugCheckout

Checkout button does nothing

“I click Pay and nothing happens.”
store.example/checkoutChrome · 1440 × 900
store.example / checkout
EVERYDAY One good thing.
On its way to you.
Contact
••••••••••••@••••••••
Payment details
••••   ••••   ••••   •••• ▰
MM / YY•••
Pay €48.00
The everyday tote€48.00

Forest / one size

Total€48.00
Illustrative screenshot · sample checkout

The screen, alongside the words. Screenshot capture depends on consent.

store.example / checkout
EVERYDAY One good thing.
On its way to you.
Contact
••••••••••••@••••••••
Payment details
••••   ••••   ••••   •••• ▰
MM / YY•••
Pay €48.00
The everyday tote€48.00

Forest / one size

Total€48.00
Simulated replay · inputs masked

A simulated example of an opt-in replay. Inputs are masked by default.

Captured console1 error
TypeError: Failed to fetch
Page
/checkout
Browser
Chrome
Viewport
1440 × 900
Language
en-GB

Captured browser context keeps the report grounded in what happened.

✳   AI summary

Checkout stalls after clicking Pay.

The user reports that the Pay button produces no visible response.

Suggested reproduction steps
  1. Open the checkout page.
  2. Click the Pay button.
  3. Check whether a confirmation appears.
AI hypothesis

Check the failed checkout request.

An inference to investigate, based on the feedback and console error.

AI interpretation stays separate from the original feedback and captured facts.

Your next move

Send the issue to your coding agent, or to GitHub or Linear.

What goes into the issue
Illustrative issue · no issue is created

Checkout button does nothing

The user reports that clicking Pay produces no visible response.

Suggested reproduction steps
  1. Open the checkout page.
  2. Click the Pay button.
  3. Check whether a confirmation appears.
View the example report

04 / AT HOME IN YOUR STACK

Small install.
Big context.

A browser SDK that fits your app. Hand the setup to your coding agent, or drop in the script tag yourself.

  1. 1

    Create an app.
    Give your feedback a place to land.

  2. 2

    Add your app’s domain.
    Add your site’s hostname under settings > install & domains > Allowed addresses.

  3. 3

    Use your public embed key.
    Copy the app’s pk_… key and initialize the SDK in the browser.

Read the quickstart
PASTE THIS INTO YOUR CODING AGENT
Add Testosaurus to this app by following https://testosaurus.dev/docs/connect.md (read it with curl -fsSL).
Your agent opens a page where you sign up or sign in, confirm your email in the same browser if asked, check the code and click Allow. It then creates your app, adds your app’s domain, installs the widget and connects itself so it can fix what testers report. Nothing secret is copied by hand. Read how connecting works.
INDEX.HTML
<script>window.Testosaurus = window.Testosaurus || { q: [] }; window.Testosaurus.init = function (o) { this.q.push(o); };</script>
<script defer src="https://testosaurus.dev/testosaurus.js?v=ecb9344fa761"></script>
<script>
  Testosaurus.init({
    endpoint: 'https://testosaurus.dev',
    apiKey: 'pk_YOUR_APP_KEY'
  });
</script>
Place once in your layout or page head; the deferred SDK never blocks rendering. Replace the example key with your app’s public embed key. Use the versioned snippet from your dashboard for production.
YOUR AGENT. YOUR RULES.

Your testers’ trust.
Your control.

Testers decide what they share. With your own agent, nothing is pushed or merged without you, and it asks before it edits.

Consent comes first

Screenshots and replay are opt-in. Replay inputs are masked by default.

Capture & consent

Set the AI budget

Run it for me (beta) takes a per-run cap you set and uses prepaid credit. Unused reserved credit returns when the run settles.

Understand billing

Your feedback to keep

JSON export of your reports on every plan, plus an API to build your own workflows.

Explore the API
SMALL START. PLENTY OF ROOM.

Start small.
Scale when you’re ready.

Start free. Add capacity when you need it.
Agent access is the same on every plan.

Free

For your next launch

€0/month

Start listening. Your agent does the fixing.

Start free
  • 1 app
  • 100 reports / month
  • Unlimited testers
  • Coding-agent access over MCP (all prompts and tools)
  • AI triage: €1/month included
  • 90-day retention

Need more capacity or specific support? Let’s talk about a custom plan

Run it for me (beta): managed coding runs use prepaid AI credit at model rates +5%, with a per-run cap you set. GitHub Actions minutes are billed by GitHub. Full pricing details

A FEW GOOD QUESTIONS

Before you
bring Testo in.

Something else on your mind?
Let’s talk

Does it work with my stack?

Yes. The browser widget works with React, Vue, Svelte, Next.js, or plain HTML. Add the script tag to your page or root layout, register your domain and use your app’s publishable key. The quickstart has the exact snippets for the Next.js app router and Vite.

Read the quickstart
What does a report actually capture?

The user’s message travels with browser context such as the page URL, viewport, language, and console entries. Screenshots and session replays are optional and require consent. The example above shows a report with these attachments enabled.

See capture and consent details
How are reports grouped and ranked?

Reports about the same problem are grouped into one shared issue, so three testers hitting the same broken button are one line in your inbox, not three. Each issue gets an AI analysis: a summary, category, severity, reproduction steps, and a suggested cause to investigate. Issues are ranked by impact: severity first, then how many reports. Counts are reports, not distinct people. The original feedback always stays available, and you can collect reports without AI. Free includes €1/month of AI credit for this analysis; Pro includes €3. Your model and report size determine how far that credit goes. On Free, AI analysis pauses when the allowance is spent.

See how reports are analyzed
Can a coding agent fix what testers report?

Yes, and it is your own agent. Connect Claude Code, Codex, or any MCP client with an access key. Run the fix_top_issue prompt and the agent picks the top-ranked issue, shows you the evidence, asks before editing, fixes it on a branch in your repo, runs your tests, and records the fix back. With your own agent, nothing is pushed or merged without you. If you would rather not drive an agent yourself, Run it for me (beta) lets a coding role in your own GitHub Actions propose a draft PR that has passed your configured tests. It is for Node.js 22 projects, needs GitHub repo and workflow authorization and prepaid AI credit, and never merges. You start each run, or opt in to automatic runs for new reports, which open a draft PR without your review.

Connect your coding agent
Can my agent set up Testosaurus for me?

Yes. Paste the one-line connect from the top of this page into your agent, or use “Connect a coding agent” under settings → agent access. Through MCP, an authorized agent can configure the app, inspect reports and evidence, and prepare coding roles. You complete GitHub and Linear connections through their normal sign-in and authorization screens. Writable access is required to change settings or start coding runs.

Read the agent setup guide
Can I control spending and take my data with me?

Yes. AI uses published model catalog token rates plus 5%, through Vercel AI Gateway. Managed coding runs reserve their maximum budget from prepaid AI credit and return the unused portion when settled. Set run and monthly coding budgets, usage limits, and alerts. On Pro, opt into additional AI report analysis using prepaid credit. Keep it funded with AI credit top-ups, optional automatic top-ups, or recurring credit on a monthly Pro subscription. JSON export of your reports is available on every plan.

Explore usage and billing
FROM REPORT TO FIX.

Your next tester report
could start the fix.

Testers report. Testosaurus ranks.
Your agent fixes and you review.

Testo waving you in