Complete SMS verification in Cursor Agent
Use Cursor Agent beside the app and tests you are editing. Install the project skill, connect the intended company number, and ask for a complete verification result.
Add a project skill Cursor can discover
npx skills add https://www.phonekit.io --skill phonekit --agent cursor --copyRun the installer in the project open in Cursor. It uses Cursor's supported .agents/skills project directory. Start a new Agent chat, check Customize → Skills, then invoke /phonekit in Agent chat. Cursor's skills documentation describes project directories and discovery. The CLI runs in the Agent's command environment, with Node.js available.
Choose a Phonekit connection
Use the CLI route below, or hosted MCP on an enabled deployment. Skill installation and inbox authorization are separate.
Connect through the CLI
npx phonekit@latest docs agents
npx phonekit@latest login --wait 50
npx phonekit@latest number listOpen the printed approval link, check its short code, and select the numbers this connection may read. Exit code 8 means approval is still pending; repeat login to resume the same request. Node.js 20.3 or newer is required. Use the quickstart to name the intended number for the project.
Connect through Cursor MCP
For an enabled hosted deployment, merge this server entry into the project's .cursor/mcp.json; preserve any other configured servers:
{
"mcpServers": {
"phonekit": {
"url": "https://www.phonekit.io/api/mcp"
}
}
}Open Customize → MCP, authenticate the Phonekit server and approve the workspace, actions and line selections in the browser. Confirm the server is connected and exposes tools. Cursor handles OAuth; this setup does not require a Phonekit CLI login or an API key pasted into the JSON. Follow Cursor's MCP guide.
The skill uses an authenticated MCP connection when available, otherwise the CLI. Public hosted setup is awaiting acceptance in Cursor; use the CLI until that check is complete. A Cloud Agent needs its own configured connection.
Debug the real user journey in place
Use /phonekit with the assigned staging number to reproduce the reported sign-in issue. Inspect the current app and existing test selectors. Mark before sending, receive the new code within 120 seconds, and finish the browser flow. Report the failing stage and the visible application state. If you change code, rerun the same journey and show the final signed-in assertion; keep the received code out of the report.
A useful debugging result
Before: the observed failing stage · Change: the relevant application or test fix · After: the same journey reaching its expected state
Prepare the actual Cloud Agent environment
A project skill travels with the repository. Personal ~/.cursor/skills can use Cursor's explicit Cloud Agents sync setting; other local compatibility directories are not automatically copied. Cloud work also needs its own Node runtime, application access and secret configuration.
Install the SDK using the current SDK installation guide before generating a test. If the task becomes a recurring test, put receipt in the SDK-based test rather than relying on the editor session. The Playwright guide is a useful starting point.
Questions you may have
Why does Customize not show Phonekit?
Check that the file is named SKILL.md under the project’s .agents/skills/phonekit directory and that Cursor has the intended repository open. Confirm discovery in Customize before treating a CLI or delivery error as a skill issue.
Does personal skill sync also copy my Phonekit login?
Skill sync supplies instructions. Configure the receiving credential and number in the Cloud Agent’s runtime separately. The local device connection is not a cloud secret distribution mechanism.
Should I use a Cursor plugin instead?
The skill installer works without a marketplace listing. A distributed Cursor plugin needs the platform’s plugin and marketplace packaging. Use that route once the Phonekit package is actually published there.
How do I turn a successful Agent run into durable coverage?
Ask Agent to put the SDK receiving step and final assertion into the repository’s test suite, then run it in CI with an explicitly scoped key and an assigned number.