Use case
I use Maestro in an environment where MCP usage is restricted, so most of my local workflow has to happen through the CLI.
For quick debugging, I often want to do small device actions without writing a full flow file first. Things like tapping a visible element, taking a screenshot, launching or stopping an app, typing into a focused field, or pressing back.
Maestro already supports these kinds of actions through flows, and some of this is available in MCP-focused workflows, but there is not a simple direct CLI version for the common one-off actions.
Proposal
Add direct CLI commands for the common device-control actions:
maestro tap-onmaestro take-screenshotmaestro launch-appmaestro stop-appmaestro input-textmaestro back
Example usage:
maestro tap-on --text "Login"
maestro take-screenshot --output artifacts/home.png
maestro launch-app com.example.app
maestro input-text "hello world"
maestro back
maestro stop-app com.example.app
The goal is to reuse Maestro's existing command/orchestration paths where possible, not add separate driver behavior for the CLI.
This would make the CLI more useful for shell scripts, local debugging, and agent-assisted workflows where MCP is not available or allowed.
Anything else?
I brought up the general CLI/MCP parity idea in last week's office hours. I have a draft implementation ready and can open a PR if this direction makes sense.
Use case
I use Maestro in an environment where MCP usage is restricted, so most of my local workflow has to happen through the CLI.
For quick debugging, I often want to do small device actions without writing a full flow file first. Things like tapping a visible element, taking a screenshot, launching or stopping an app, typing into a focused field, or pressing back.
Maestro already supports these kinds of actions through flows, and some of this is available in MCP-focused workflows, but there is not a simple direct CLI version for the common one-off actions.
Proposal
Add direct CLI commands for the common device-control actions:
maestro tap-onmaestro take-screenshotmaestro launch-appmaestro stop-appmaestro input-textmaestro backExample usage:
The goal is to reuse Maestro's existing command/orchestration paths where possible, not add separate driver behavior for the CLI.
This would make the CLI more useful for shell scripts, local debugging, and agent-assisted workflows where MCP is not available or allowed.
Anything else?
I brought up the general CLI/MCP parity idea in last week's office hours. I have a draft implementation ready and can open a PR if this direction makes sense.