CertiQA Academy
Learn test automation and AI agent testing.
Practical guides for QA engineers and developers: how to move from manual testing to automation, and how to test AI agents that call tools and change real systems. Free to read, and written to work with any tool.
Guides
Pick a starting point, or read them in order.
Foundations
Test automation foundations
The shift from manual to automated testing: locators, waiting, assertions, test data and how a manual test case becomes an automated one.
- Locator strategies and why some break
- Waiting on conditions instead of sleeping
- Setup, seed data and repeatable tests
Reference
QA and agent testing glossary
Plain-English definitions from fixtures and flaky tests to tool mocking, LLM judges and MCP, arranged A to Z.
- Classic automation terms
- Agent, tool and judge vocabulary
- Security and release terms
Agent testing
AI agent testing 101
Why classical assertions fail on agents and what to check instead: tool calls, step budgets, judged answers, attacks, and the whole customer journey.
- The agent loop and where to test it
- A worked refund-agent example
- Testing past the reply, into your systems
Checklist
Agent pre-flight checklist
Twenty checks to run before releasing an agent, in five groups, with ticks you can work through or print.
- Tool isolation and failure handling
- Limits, cost and loops
- Attacks, grounding and release gates
Career
QA career transition guide
A 30-60-90 day plan for moving from manual QA toward automation and agent testing, with portfolio ideas and interview topics.
- How the role changes at each stage
- A staged learning plan
- Portfolio projects to build
How the work changes
From manual testing to testing agents.
A rough picture of how responsibilities shift as testers take on automation and then agents. Real roles vary, so treat it as a map and not a ladder.
| Manual and functional QA | Test automation | Agent and AI quality | |
|---|---|---|---|
| Main focus | Exploratory testing, reproducing bugs, manual regression | Automated regression and CI test gates | Verifying non-deterministic agents, tool use and prompt safety |
| What is tested | Web forms, standard REST APIs, navigation | Single-page apps and service APIs | Multi-turn agents, tool-calling loops, tool-call interception |
| How results are checked | A person looks at the screen and text | Exact assertions on page state and responses | Tool-call checks, rubric-scored meaning, step limits |
| Typical pain | Repeating regression every sprint | Brittle selectors and maintenance | Hallucination, runaway loops, drift, prompt injection |
| Release evidence | A sign-off sheet or ticket comment | A green CI run and a test report | Pass rates over repeated runs, attack results, a release report |
Practice what you learn
Try the ideas on a real application.
Reading helps, but tests make the concepts stick. CertiQA lets you practice without setting up a framework first.
- Click elements in a live browser and see the selectors it suggests.
- Assemble reusable steps, such as login or checkout, and reuse them across tests.
- Export a Playwright script when you want to read the code.
- Mock an agent's tools, inject a failure and run an attack against it.
1 Open /cart
2 Click "Pay with card"
3 Mock API payments -> DECLINED
4 Assert error message is visible
5 Assert order status is still PENDINGFAQ
Common questions from testers learning automation
Can a manual tester learn test automation?
Yes. Start with the concepts rather than a framework: how a page is structured, what makes a locator stable, and how an assertion proves behavior. A visual recorder or reusable blocks help you practice those ideas before you write much code, and the code follows once the concepts are clear.
Start with the foundationsDo I need to know how to code to start?
Not to begin. Many teams build first tests from recorded or plain-step flows. Reading and writing code, such as TypeScript with Playwright, becomes important as your suite grows, and it is much easier once you understand what a test is supposed to prove.
What is the difference between test automation and AI agent testing?
Automation checks deterministic software, where the same action gives the same result. Agent testing checks systems whose wording, reasoning and tool choices change between runs, so it checks tool calls, limits, resulting state and rubric-scored answers, and measures a pass rate over repeated runs.
Read agent testing 101Why learn agent testing if my company only has regular web apps?
Many products are adding assistants, support bots and automated workflows, and someone has to say whether they are safe to release. Knowing how to test tool calls, set limits and run attack cases is a transferable skill even if your current product has no agent yet.
Which tools should I learn first?
For web UI testing, Playwright is a widely used open-source choice with built-in waiting. For agents, learn the Model Context Protocol and how LLM judges and tool mocking work. The concepts carry across tools, so learn those first.
Is the Academy free?
Yes. The guides are free to read, and you do not need an account. If you want to practice, the CertiQA 14-day trial needs no card.
Also see example scenarios, how CertiQA compares and authoring tests with an AI assistant.
Learn it, then test your own agent.
Start with one conversation that worries you and see how your agent behaves.