mobium check
This tool on every surface: CLI · MCP · Python · JavaScript · Go · Java · .NET
Check a claim about a recorded session, with an independent model
A model with a fresh context reads the recording — each call, its result or error, and the screen and screenshot after it — through read-only tools, and answers PASSED, FAILED or INCONCLUSIVE with the observations it rests on.
Inconclusive means the recording cannot settle the claim. A model's judgment
is not proof: keep wait and test assertions for what can be asserted
directly. Needs no device. The default judge, Claude, needs an Anthropic
credential (ant auth login, or ANTHROPIC_API_KEY where the daemon starts),
and each check is a paid call; it is built but not yet measured against
claims with known answers. openai:recording with verdict lets an agent you run be the judge.
See docs/guides/verifier.md. A check against a live device is not built yet.
Usage
mobium check <claim> -i <recording.zip> [flags]
Flags
| Flag | Type | Default | What it does |
|---|---|---|---|
--input -i | string | The recording: the .zip that mobium trace stop saved | |
--model | string | The model that judges (default claude-opus-5-5, not yet measured); openai: |
And the global flags.
Examples
Ask a model on this machine whether a claim holds in a recording
mobium check "After card 3 was tapped, it was shown face up as the Blue Morpho." --input card.zip --model openai:qwen2.5:14b
More, from mobium check --help
mobium trace start && mobium set testid=termsCheck && mobium trace stop -o form.zip
mobium check "the terms box was ticked after the tap" -i form.zip
No screenshot: it reads a recording, not a device, and needs no screen at all.
The tool
check calls app_check; its arguments and what every client calls it are there.