← All plugins

AI Assistant plugin · Integrations

Your AI agent, with a real device in its hands.

The AI Assistant plugin connects AI agents to Dev Scenario. The agent sees the screen on the device, writes the test, runs it, reads what failed and fixes it, until it passes. Use the assistant built into the app, or Claude Code, Cursor or GitHub Copilot.

Pro and Team
What it does

Everything in the AI Assistant tab.

See the screen

The current screen as text the model can read, a screenshot, and selectors checked against the live UI.

Write and validate

The full command reference and the real parser, so the agent checks a flow before it runs it.

Run on a device

Flows, scenarios and chains on the selected emulator, simulator or phone, with the result back in the chat.

Understand failures

The failing step with its screenshot, failed requests, device logs and the test's history.

Save in the right place

The agent reads your project's structure, reuses shared steps like login and saves in your naming style.

You approve risky calls

By default, anything that reaches past the app, like shell commands, installs or secrets, waits for your approval first.

Use cases

How teams use it.

New tests

Write a test from one sentence

“Test that a user can change their password.” The agent explores the screens, writes the flow and runs it until it passes.

Hours of test writing become minutes of review.

Red runs

Triage last night's failures

“Why did checkout fail?” It reads the report, the failed calls and the logs, and says whether the app or the test broke.

Know what to fix before standup.

Backend states

Mock by asking

“Make the cart endpoint return an empty list and check the empty state.” The agent adds the mock and runs the check.

Edge cases without hand-written mocks.

Upkeep

Fix tests after a UI change

A screen changed and five flows broke. The agent looks at the new screen, updates the selectors and reruns them.

Test upkeep stops eating the sprint.

Good to know
  • Included in Pro and Team, and in every plan's 6-week trial. See pricing.
  • Works with any tool that supports MCP; connecting Claude Code, GitHub Copilot CLI, Codex or Gemini CLI takes one click in the app.
  • The built-in assistant uses your own Claude Code or GitHub Copilot sign-in.
  • More on the workflow: Dev Scenario for AI agents.