QAFlow Agent

Guided QA lifecycle automation with Gemini, Google ADK, and MongoDB MCP

MongoDB Track · Human-in-the-loop QA

1. Requirement Intake

Enter a real requirement, or load the sample requirement for demo. QAFlow will run the workflow and pause at approval gates.
Use Gemini 2.5 Flash for stable demos. Gemini 3 options are kept for hackathon alignment, but they require access in your Google Cloud project/region.

2. Approval Gates

Approve only after reviewing the visible outputs. QAFlow will not continue through governance gates automatically.

Requirement rejected. Update the requirement fields above, then resubmit it for another RAG-based gap analysis.

3. Human Review Recommendations

QAFlow can suggest editable updates for human reviewers. BA/PO can refine requirement clarifications, and QA Lead/Manager can refine test case rework notes before resubmitting.

BA / Product Owner Recommendation

Visible after requirement gap analysis or requirement rejection. Edit this recommendation, apply it to Acceptance Criteria, then resubmit the requirement.

QA Lead / Manager Recommendation

Visible after test case peer review or test case rejection. Edit the recommendation, then regenerate test cases with these rework notes.

4. Test Execution Input

Choose simulated execution for demo, or enter manual execution results from QA/automation runs before running Test Execution.

5. Manual Advanced Actions

Optional. Actions are locked until prerequisite steps are complete.