End-to-end tests, written by your coding agent

Beryl is the web testing integration for your coding agent. Your agent writes tests in plain language; Beryl runs and heals them so regressions never reach your users.

What we do

Tests authored from your words

Beryl plugs into Claude Code, Cursor, VS Code, opencode and other coding agents over MCP. Hand it a document describing your flows and Beryl's cloud agent explores your app in a real browser and authors readable end-to-end tests as plain-language plans, proving every save with a replay. No selectors to maintain, and every test exports as standard Playwright.

Reliability on autopilot

Beryl runs your tests in real cloud browsers, on a schedule, or as a CI gate. Cosmetic drift is healed from each test's recorded intent, real breaks stay red, and Slack hears about them before your users do.

The people behind it

Ankesh Khemani

Founder and CEO

Former Principal Engineer and Engineering Manager at Atlassian, where he built and scaled platform infrastructure serving thousands of engineers. Ankesh understands the reliability problem from the inside, not as a consultant, but as someone who owned the pager, ran the post-mortems, and felt firsthand the gap between what testing tools promise and what engineering teams actually need.

Engineering team

Tushar Singh

Software Engineer

Akshat Gupta

Software Engineer

Tarpita Singh

Software Engineer

Tanay Agarwal

Software Engineer