Add your evidence
Upload up to five screenshots, paste or load a log file, and write a short note about what went wrong. Add the app name and browser if you know them.
Early access · bring your own Anthropic API key
Amnezi is a small web app for QA engineers and developers. Drop in what you have, review what Claude drafts, and see which piece of evidence backs each step.
No account. No server-side storage. You use your own Anthropic API key, so API usage is billed to your Anthropic account.
How it works
Upload up to five screenshots, paste or load a log file, and write a short note about what went wrong. Add the app name and browser if you know them.
Before anything is sent, Amnezi masks emails, API keys, tokens, IP addresses and similar patterns in text, and shows you what it masked.
You get numbered reproduction steps, each pointing to the log lines or screenshot it relies on, plus open questions and a Playwright test draft to review.
Sample
A made-up checkout bug, to show the shape of the result. Every step carries an evidence reference.
Sample output (illustrative, not a customer report)
Does the freeze also happen with a newly entered card, or only with saved payment methods? The evidence does not say.
import { test, expect } from '@playwright/test';
test('checkout completes with a saved card', async ({ page }) => {
await page.goto('/cart');
await page.getByRole('button', { name: 'Checkout' }).click();
await page.getByLabel('Saved card').check();
await page.getByRole('button', { name: 'Pay now' }).click();
await expect(page.getByText('Order confirmed')).toBeVisible();
});
Why Claude
Claude reads error dialogs, UI state and layout from the images you upload, so a screenshot is evidence and not just an attachment.
Real logs are long. Claude can take a large log (Amnezi accepts up to 200 KB of text) in one request and relate a line near the end to what the screenshot shows.
Amnezi forces a tool call with a fixed schema, so the result is steps, evidence references, confidence and questions, not free-form prose.
A model can still cite something that is not there. Amnezi validates every evidence reference against the material you provided and drops any reference that points outside it (a log line that does not exist, a screenshot you did not upload).
Privacy
The Anthropic API key is used from the page. It is kept only in sessionStorage, and only if you tick "remember for this tab". Closing the tab clears it.
Requests go from your browser straight to the Anthropic API under your own account. Amnezi has no server that receives your screenshots, logs or notes.
Text is scanned in the browser first and common secrets are masked. This is pattern-based and best-effort, so review the preview. Screenshots are sent as you upload them; crop or blur anything sensitive first.
There is nothing to sign up for. The site sets no cookies and runs no analytics. Read the full privacy notice.
Roadmap
Next and Later are plans and may change. The source is MIT-licensed on GitHub.
Company
Amnezi is developed by Amnezi Teknoloji Limited Şirketi, a limited company registered in Türkiye and founded in January 2026.
Founder: Eren Çivril · GitHub profile
Contact: erencivril@amnezi.dev · GitHub
We are looking for feedback from QA engineers and developers who deal with unclear bug reports. Please do not email sensitive logs or customer data.
erencivril@amnezi.dev