Approxima provides an AI‑driven testing agent that automatically runs end‑to‑end checks on every pull‑request preview deployment. It detects UI regressions by parsing the PR diff, executing targeted interactions in a real browser sandbox, and posts screenshots and explanations as GitHub comments. The service integrates with CI/CD platforms and supports declarative “journey” tests without requiring custom test code.
Funding
Funding not disclosed
Founders
Product
Problem
Developers rely on static code analysis and manual QA to catch regressions, but these approaches miss runtime bugs that only appear in a live browser environment. As a result, functional, visual, or integration defects often reach code review or staging, causing delays and additional rework.
Solution
Approxima delivers an AI‑driven testing agent that automatically validates every pull‑request preview deployment. The agent parses the PR diff, infers the affected UI paths, and executes targeted end‑to‑end interactions in a real browser sandbox. When a failure is detected, it captures a screenshot, generates a concise explanation, and posts a pass/fail verdict directly on the pull request. The service plugs into existing CI/CD platforms (Vercel, Netlify, Cloudflare, etc.) and requires no test scripts or custom configuration. Teams can also author reusable “journeys”—plain‑language test flows that run on each PR or on a schedule—to extend coverage beyond the diff‑generated checks.
Target Audience
The primary customers are engineering teams building modern web applications that use preview deployments and automated CI/CD pipelines, including front‑end developers, QA engineers, and product managers seeking continuous reliability.
Features
- Diff‑aware AI engine that extracts changed components and creates focused validation flows without manual test code
- Real‑browser execution in an isolated, ephemerally provisioned container, supporting Chrome‑compatible rendering and mobile viewports
- Automatic capture of screenshots, error logs, and human‑readable explanations posted as GitHub PR comments with status checks
- Declarative “journey” editor allowing non‑technical definition of multi‑step user scenarios that run on every PR or on a cron schedule
- Native integration with Vercel, Netlify, Cloudflare and any platform that publishes preview URLs, using minimal GitHub App permissions (read code, write comments)
- Framework‑agnostic support for React, Next.js, Vue, Nuxt, Angular, Svelte, Remix, Astro, and plain HTML applications
- Security‑first sandboxing: source code is never persisted; only test artifacts (screenshots, logs) are retained after each run
- Usage‑based billing model that scales with the number of PR validations executed