Skip to main content

​
Start from where you work

GitHubGet a verdict from a live session on every pull requestCoding agentLet Cursor, Claude Code, Codex, or another agent drive your appCLIStart sessions, check what your app does, and query Atlas from your terminalDashboardOpen Atlas, sessions, and reports in the browser

​
How it works

Every change goes through the same loop. Each step is something you can open in Revyl.
  1. Intent. A pull request, a prompt to your coding agent, or the plain-English steps of a test.
  2. Agent execution. Your coding agent or the Revyl agent runs the app in a cloud session: a proof run, a test run, or a dev loop.
  3. Runtime evidence. Each run leaves a report with video, screenshots, logs, and network traffic.
  4. Verified behavior. Validation steps and the proof verdict record whether the app did what was intended.
  5. Stored behavior. Atlas adds the screens, pages, and transitions the run observed to your app’s record.
  6. Questions. People open Atlas in the dashboard. Agents query it with revyl atlas and the Atlas skill.
  7. New intent. What you learn, such as a screen no test covers, becomes the next change, test, or exploration.

​
What Revyl handles

​
Keep going

Core concepts

Runtimes, evidence, Atlas, and the objects behind them

Frameworks

Build setup for Web (alpha), iOS, Android, React Native, Expo, Flutter, KMP, and Bazel

Library

Modules, variables, scripts, and files you reuse across tests

Integrations

GitHub, GitLab, Slack, Expo, Cursor, and Capy

Security and network

IP allowlists and AWS PrivateLink

CLI reference

Every revyl command and flag