Typed from the start
Strict domain contracts keep stories, plans, tests, evidence, and reports structured and inspectable.
QualityWorkItem✓ valid
Governed AI quality engineering
Let AI move fast. Keep quality in control.
QUALTAN turns requirements into validated test artifacts through typed contracts, deterministic gates, and explicit human approval.
Open source · Apache 2.0 · Python 3.11+
MODEL OUTPUTProposed✓
HUMAN GATEApproved✓
typed contracts
audit trail
Approval-gatedExecution and mutation controls
DeterministicValidation before progression
Evidence-firstDurable artifacts and telemetry
Local-firstOffline demo, no surprise calls
01 / The reason
AI can propose the answer. Your team still owns the decision.
QUALTAN separates probabilistic model reasoning from the deterministic systems that validate, approve, and record what happens next.
Validated before it moves
Deterministic quality gates catch missing coverage, unsafe source, schema drift, and compile failures.
Approval is a feature
Durable workflows pause at the right moment, preserving the review decision and the evidence behind it.
review required→human
02 / The capability map
One governed control plane for the quality lifecycle.
Start with a requirement and finish with a reviewable, reproducible record of what was proposed, validated, approved, and run.
Jira and API intake
Bring in stories, acceptance criteria, OpenAPI descriptions, or approved quality knowledge.
Inspect the boundary ↗Risk-based planning
Analyze intent, risk, coverage, test data, security, and performance needs through typed services.
See the workflow ↗Test generation
Produce bounded Playwright and quality artifacts that are checked for source safety and compilation.
View the quality gates ↗Security and performance
Keep sensitive data redacted and translate approved plans into safe, bounded security and Locust scenarios.
Explore the agents ↗Jira, X-Ray, and MCP
Connect through narrow, auditable adapters. External mutations remain disabled until explicitly enabled.
Read the MCP surface ↗Evidence that travels
Retain reports, traces, approvals, telemetry, and evaluation results for review and handover.
See the infrastructure ↗03 / The workflow
From requirement to evidence, with no mystery steps.
01Ingest
Bring in a story, spec, or API.
02Analyze
Map intent, risk, and coverage.
03Generate
Propose an executable plan.
04Approve
Keep a human in the loop.
05Validate
Run gates and capture proof.
Every meaningful node is checkpointed. Stop, review, and resume without losing the thread.
Read the framework04 / The boundary
Built for teams who want leverage without surrendering judgment.
01
Models propose.
AI handles bounded extraction, reasoning, generation, and diagnosis.
02
Policies decide.
Execution hosts, external mutations, and sensitive data stay behind explicit controls.
03
Evidence remains.
Artifacts, approvals, telemetry, and outcomes form a durable audit trail.
04
Production scales deliberately.
Keep the control plane stateless, move long-running work to isolated workers, and externalize durable state before multi-replica operation.
05 / The first ten minutes
Try the safe path before you connect a live system.
01
Install
Use Python 3.11+ and install the package with its test extras.
pip install "qualtan[test]"02
Run offline
Exercise the typed workflow with deterministic fixtures and no model or Jira account.
pytest -q tests/test_modernized_framework.py03
Connect carefully
Configure approved providers only after reviewing policy posture and integration boundaries.
Open the offline demo ↗Ready when you are