MCP Launch Video
How it works
1
Point your agent to the failed run
Tell your agent about the failure: “Help me fix the bug from this failed test” or paste a run ID directly. You can give the shareable link (the agent uses its
public_id), or the direct link to the test run (task_id).The agent calls get_test_run with include: ["overview"] to identify which steps failed and why.2
Drill into the details
The agent pulls deeper context in one call —
get_test_run(include: ["steps", "console", "network"]) — to surface JavaScript errors, failed HTTP requests, or unexpected behavior. It uses contains on the log sections to grep for the relevant lines rather than pulling whole logs.3
Inspect screenshots
If the agent deems it necessary, or you provide that instruction, it retrieves screenshots via
get_test_run(include: ["screenshots"], stepRef: "Step 6.14") — stepRef uses the label the app shows for one result. Pass position: "before" | "after" | "both" to pick the frame.4
Fix the issue
With the full context available, your agent can help you identify the root cause and suggest or apply a fix directly in your codebase.
