Skip to main content
This journey shows a complete path from setup to productive agent-driven verification.

​
1. Install and authenticate

brew install RevylAI/tap/revyl
revyl auth login
revyl auth status
If you use npm or pip, see CLI installation options.

​
2. Configure MCP for your coding tool

Use one of these:
  1. Codex: codex mcp add revyl -- revyl mcp serve
  2. Claude Code: claude mcp add revyl -- revyl mcp serve
  3. Cursor: use the one-click install on MCP Setup or add revyl mcp serve in Cursor MCP settings.
Then verify locally: 
revyl mcp serve

​
3. Install the Revyl skill

revyl skill install
By default this installs the CLI skill family. For MCP-driven agent flows, install MCP skills explicitly: 
revyl skill install --mcp
If you want one tool explicitly: 
revyl skill install --codex
revyl skill install --cursor
revyl skill install --claude
revyl skill install --codex --mcp
revyl skill revyl-mcp-dev-loop install --codex
Restart your IDE/agent after skill installation.

​
4. Start your dev loop

revyl init
revyl init --hotreload
git checkout -b feature/new-login
revyl build upload --platform ios-dev
revyl dev
When ready, Revyl prints a viewer URL and deep link details. If you already have a local artifact and want to skip the build command: 
revyl build upload --platform ios-dev --skip-build
revyl dev --platform ios

​
5. Give the agent a high-leverage prompt

Use this template: 
Use Revyl MCP tools only.
Goal: bypass login and reach the home screen.
For each step: screenshot, briefly describe what is visible, take one best action, then screenshot again to verify.
If anything unexpected happens, re-observe before taking another action.
At the end, summarize final screen, actions taken, and bugs found.

​
6. Re-anchor if the agent starts acting blind

Pause actions. Re-anchor now:
1) screenshot,
2) describe current screen,
3) choose one action with reason,
4) screenshot verify.
Then continue.

​
7. Close the loop cleanly

  1. Stop the dev session with Ctrl+C in the revyl dev terminal.
  2. Ask the agent for a short execution summary.
  3. Convert successful ad hoc flows to reusable tests.
Continue with Agent Journey: Ad Hoc to Test.