Debugging mobile test failures
When a test fails, Semaloop gives you a high-FPS video replay, a step-by-step interaction timeline, system logs, network activity, a root cause analysis and a suggested fix, all on the run's page.
What you get
- High-FPS video replay, with audio where the test used the microphone or played media
- An interaction timeline showing every step and the agent's reasoning
- System logs and network activity, with logs downloadable from the dashboard
- A root cause analysis and a suggested fix on every failure
- A ready-made prompt that hands the run to your coding agent over MCP
- Build summaries and questions in natural language across runs, for failures that span builds
What you see on every run
A run ends in one of three outcomes. Success means the agent verified the behaviour. Failure means it could not verify that the functionality worked as expected. Error means the run hit a problem on Semaloop's side, such as a model provider outage, and the Semaloop team is told.
- Video replay: a high-FPS recording of the whole run
- Interaction timeline: a step-by-step trace with the reasoning behind each step
- Audio capture: any microphone input or media playback is included in the video
- System logs: what your app output during the run
- Network activity: responses, headers and timings
- Issues: friction the agent found along the way, such as a slow page or text that is cut off
From failure to cause
On a failure, Semaloop combines everything its agent had, including the system and network logs, into a short root cause analysis and a suggested fix. Start there, then check it against the evidence.
Step through the trace with j and k, and play or pause the video with the space bar, to see exactly which step went wrong and what was on screen. Then open the system logs and network activity to confirm what happened underneath, for example a request that returned an error. Both logs can be downloaded.
Fixing it with your coding agent
Every failure has a Send to agent prompt containing the run's ID. With the Semaloop MCP server connected, a prompt such as 'Help me fix Semaloop test run' followed by that ID lets your coding agent pull the run, the video, the network logs and the system logs, work out whether the app broke or the test was brittle, and then push a code fix or revise the spec in the same session.
Failures that span builds
Once every run on a build has finished, a written build summary says what changed since the last build, with each change linked to its run. Ask answers questions in natural language about an app's builds, tests and runs, such as which tests started failing or whether a failure only happens on one device.
Ask works from results and root cause analyses. It cannot see a run's video, trace, logs or network traffic, so it points you to the run's page for those. Recurring problems are grouped into issues across runs.
Limits worth knowing
A root cause analysis is a summary and a suggestion, not a guarantee, so treat it as a strong starting point. If a result looks wrong, answer 'Was this result correct?' on the run page. The answer is used to train the agents, with no effect on that run.
For UK teams
Run videos, logs and network traces can contain data from your test accounts. Semaloop holds SOC 2 Type II certification, and a data processing agreement is published at /dpa.
Frequently asked questions
Last reviewed: 26 October 2026
How do I debug a failed mobile test in Semaloop?
Open the run. Read the root cause analysis and suggested fix, step through the interaction timeline against the video replay, then confirm in the system logs and network activity. Or send the run to your coding agent with the Send to agent prompt.
What does a Semaloop failure include?
A high-FPS video, an interaction timeline with reasoning, system logs, network activity, audio capture where relevant, any issues the agent found, a root cause analysis and a suggested fix.
Can I download the logs?
Yes. Both system logs and network logs can be downloaded from the Semaloop dashboard.
Can a coding agent fix a failure for me?
It can start the work. Connected to the MCP server, an agent can pull the run's video, logs and root cause, decide whether the app or the test is at fault, and propose a code change or a revised spec. You review and merge as usual.
What is the difference between a failure and an error?
A failure means the agent could not verify that your functionality worked as expected. An error means the run hit a problem on Semaloop's side, such as a model provider outage, and the team is informed.
Can I ask questions about my results?
Yes. Ask answers questions in natural language about an app's builds, tests and runs, and links to the runs it mentions. It cannot see a run's video, trace, logs or network traffic.