Running

Coding agents

TestedOK as an MCP server for Claude Code, Cursor and the rest.

TestedOK is an MCP server. A coding agent that just wrote a screen can write the flow for it, run it, and read the failing step, without leaving the editor.

Setup

claude mcp add testedok -- testedok mcp

For Cursor, Windsurf or any MCP client, the command is testedok with the argument mcp, over stdio. Sign in first with testedok login on that machine; the server uses the saved token.

The dashboard's Agent setup page has a prompt to paste into the agent's instructions (or CLAUDE.md) that explains the flow format, where flows live, and when to run them.

Tools

Tool	What it does
`run_tests`	Runs a flow or a folder. Options: `simulator`, `replay`, `env`, `video`. Reports pass/fail with the failing step and a link to the run. Sends a progress notification per step, so an agent that listens sees a failure as it happens.
`get_run`	A finished run's step timeline: each action, who decided it, and what was on screen.
`list_tests`	Flows in the working directory.
`list_devices`	Booted simulators and attached Android devices.
`install_app`	Installs a built `.app` or `.apk` on a device.
`list_environments`	Environments from `testedok.yaml`.
`create_inbox`, `wait_for_code`	An email address for a sign-up, and the code that arrives.
`lease_account`, `release_account`	A test account from the project's pool, and giving it back.

What a good agent loop looks like

Build the screen.
Write tests/<flow>.json with a goal in plain English and expectText for what the screen shows when it worked.
run_tests. On failure, read the step and the screen; decide whether the flow's wording or the app is wrong. Fix the one that is wrong.
Commit the flow with the code.

Runs from an agent are marked as such in the dashboard, so you can see what it did while you were away.