The AI-powered QA operating system for any project, any tech stack. Built on GitHub Copilot. Deployed in 30 minutes. Scales to 1000+ engineers.
QAForge is a multi-agent AI orchestration system that automates the full QA lifecycle — from reading a Jira story to delivering running automated tests and audit-ready reports.
You paste a story. Eight specialized AI agents collaborate, with your approval at every step, to:
- Analyse the story and identify every test scenario
- Write structured BDD/Gherkin test cases (Xray-ready)
- Resolve real, live test data for each scenario
- Generate production-ready automated test code
- Validate coverage against every acceptance criterion
- Report results with a final stakeholder report + Jira import
No manual handoffs. No context switching. No starting from scratch every sprint.
🎥 Watch the full 25-minute live demo on YouTube — end-to-end pipeline run on a live site, no cuts.
You: @QAForge Manager
[paste your Jira story]
Manager: Story understood. Folder created at FeatureSpecs/TICKET-123/.
Delegating to Story Analyst...
Analyst: Analysis complete. 14 scenarios identified across 4 test suites.
Saved to 02_TEST_ANALYSIS.md.
⏸ APPROVAL GATE — type APPROVE to continue.
You: APPROVE
Generator: 23 BDD test cases written in Gherkin format.
P0: 8 cases | P1: 11 cases | P2: 4 cases.
⏸ APPROVAL GATE
You: APPROVE
Data Resolver: Live API called. 5 test scenarios resolved with real event IDs.
Saved to 04_TEST_DATA.md.
⏸ APPROVAL GATE
You: APPROVE
Automation Engineer: 5 Playwright spec files generated.
POM updated with 3 new methods.
All fixtures wired to resolved test data.
⏸ APPROVAL GATE
You: APPROVE → run the tests → come back
Validator: 23/23 test cases mapped to ACs. 21 passing. 2 failing (P1 — non-blocking).
QA Sign-Off: CONDITIONAL PASS.
Reporter: Final report saved. Xray XML ready for Jira import.
✅ COMPLETE.
| The old way | QAForge |
|---|---|
| Write test analysis manually (2–4 hours) | Analyst agent — 2 minutes |
| Write test cases from scratch every story | Test Case Generator — 3 minutes |
| Chase down test data per environment | Data Resolver — automated, live data |
| Write Playwright/Selenium code by hand | Automation Engineer — generated, fits existing framework |
| Validation is a mental checklist | Validator — every AC mapped, every gap flagged |
| Report is a copy-paste from Jira | Reporter — structured, audit-ready, Xray-importable |
| Each engineer does this differently | QAForge — consistent across every engineer, every story |
- QA Engineers who want to stop writing the same analysis and test case documents sprint after sprint
- SDET Engineers who want AI to generate the boilerplate automation code while they focus on framework design
- QA Leads who want every engineer on the team producing consistent, traceable artifacts
- DevOps / Engineering Managers who need audit trails, Xray imports, and sign-off documents without chasing QA
QAForge ships with four skill packs covering the most common testing environments:
| Tech Stack | Use Case | Skill Pack |
|---|---|---|
| Playwright + TypeScript | Web UI testing | skills/playwright |
| REST API (any language) | API contract + integration testing | skills/api-rest |
| PostgreSQL / MySQL / SQL Server | Direct DB verification + ETL validation | skills/database-sql |
| AWS SQS + S3 + Lambda + DynamoDB | Cloud pipeline testing | skills/aws-cloud |
The Core Engine never changes. The Skill Layer adapts to your stack. The Context Layer teaches the agents your specific project.
- VS Code (latest stable)
- GitHub Copilot extension (active subscription)
- Node.js >= 18
# 1. Clone QAForge — this becomes your QA workspace
git clone https://github.com/mdrasul/qaforge.git
cd qaforge
# 2. Open in VS Code — all 16 agents are auto-discovered immediately
code .
# 3. Copy the context template and fill it in for your project
cp core/templates/PROJECT_CONTEXT.template.md PROJECT_CONTEXT.md
# Open PROJECT_CONTEXT.md and fill in your app details
# Reload VS Code: Cmd+Shift+P → Developer: Reload Window- Open Copilot Chat in VS Code (
Cmd+Shift+I) - Fill in
core/templates/INPUT_STORY_TEMPLATE.mdwith your story - Type
@QAForge Managerand paste your filled template - Follow the approval gates
Full guide: docs/getting-started.md | Troubleshooting: docs/troubleshooting.md
qaforge/
├── .github/agents/ ← All agents — auto-discovered by VS Code on clone
├── core/ ← Templates and engine reference docs
├── skills/ ← Skill pack docs and code templates
├── examples/ ← Reference implementations per stack
└── docs/ ← Full documentation
See core/README.md for agent details and docs/ for full documentation.
┌─────────────────────────────────────────┐
│ CONTEXT LAYER — Your PROJECT_CONTEXT │ ← You write this once
│ (domain knowledge, URLs, schemas) │
├─────────────────────────────────────────┤
│ SKILL LAYER — Your Tech Stack │ ← Pick from the skills library
│ (playwright / api-rest / database-sql / aws-cloud) │
├─────────────────────────────────────────┤
│ CORE ENGINE — Universal Agents │ ← Never changes
│ (Manager, Analyst, TC Gen, etc.) │
└─────────────────────────────────────────┘
Read the full design: ARCHITECTURE.md
QAForge does not run autonomously end-to-end. Every major step produces an artifact and pauses for your review.
This is intentional. AI moves fast. Judgment makes it right.
You are always in control. The agents eliminate the writing work. You provide the thinking.
QAForge is designed to grow through community-contributed skill packs.
If you've adapted QAForge for a new tech stack (GraphQL, Kafka, mobile, Kubernetes, SAP, etc.), contribute it back as a Skill Pack. See CONTRIBUTING.md.
The more skill packs exist, the more stacks QAForge covers — without changing the core engine.
MIT — free to use, modify, and distribute. See LICENSE.
MD Rasul — Senior SDET / Automation Architect 10+ years building production QA automation systems across enterprise, cloud, and web platforms.
"I built this because I was tired of watching talented QA engineers spend 80% of their time writing documents instead of solving testing problems. QAForge gives that time back."