AI Testing & MCP

Generate Test Cases From Your Requirements

Generic AI writes generic tests: plausible steps for software it has never seen. An agent on the BesTest MCP reads your actual requirements, acceptance criteria, and coverage before it writes a word. That difference is the whole point.

The blank page after every requirement

Every requirement eventually needs test cases, and drafting them is the same translation work each time: read the acceptance criteria, imagine the paths, write the steps. Skip it and coverage gaps pile up; rush it and the tests check the happy path only.

ChatGPT can draft test cases, but it drafts them blind: it has never read your requirement, your conventions, or what is already covered. Connected over MCP, the agent drafts from your actual data and files the results as linked, reviewable drafts.

Drafting blind

  • Read the requirement, stare at the blank test case form
  • Paste text into a chatbot, copy results back by hand, unlinked
  • Get plausible steps that ignore what is already covered
  • Coverage gaps wait for whoever has a free afternoon

Drafting grounded

  • The agent reads the requirement and its acceptance criteria first
  • It checks existing coverage, so it drafts what is missing, not duplicates
  • Drafts land in BesTest as linked test cases, ready for review
  • Your review workflow, with its review audit trail, keeps humans in charge
Try it yourself

One prompt, start to finish

you ask your agent
For every requirement in the payments folder with no linked test case, draft one from its acceptance criteria, link it to the requirement, and leave it in draft for review.
  1. 1

    Finds the real gaps

    It lists the requirements in scope and checks their coverage over the MCP, so it drafts for what is actually missing instead of duplicating existing work.

  2. 2

    Drafts from your requirements, not from vibes

    Steps and expected results come from each requirement’s acceptance criteria, and significance guides depth: your riskiest requirements deserve more than a happy path.

  3. 3

    Files drafts for humans to judge

    Every generated test case lands linked to its requirement, in draft, routed through the built-in review workflow. The review audit trail records who reviewed and approved what.

The result

Drafting stops being the bottleneck: gaps get filled the day they appear, every draft arrives linked and reviewable, and your test library grows in your conventions instead of a chatbot’s.

BesTest · REQ-204 after drafting
TC-201Expired card at checkoutdraft · linked
TC-202Partial refund after capturedraft · linked
TC-203Network drop mid-paymentdraft · linked
REQ-204 coverage84%

Drafts land in review, in your folder conventions, linked to the requirement they came from.

More prompts to steal

You

Draft edge-case tests for REQ-88: think beyond the happy path and list your assumptions.

You

Turn this bug report into a regression test case and link it to the related requirement.

You

Review the payments folder: which existing test cases look outdated against their requirements?

Scoped to your permissions, granted per user

Your agent acts as you: everything it reads or writes stays inside your Jira permissions. The MCP is in closed beta while we harden it, with GA coming. The request form asks for a few details from BesTest’s Project Settings in your Jira, which is what lets us enable exactly your site and user.

Request beta access

Frequently asked questions

Will it flood us with low-quality tests?

It drafts into your normal review gate, not into your runs: with reviews set to required, nothing counts until a human approves it. Generation makes drafts cheap; the review workflow keeps the bar exactly where you set it.

How is this better than pasting into ChatGPT?

Context and filing. The agent reads the actual requirement, the related test cases, and current coverage before drafting, and the output lands linked in BesTest instead of in a chat window you copy from. Same model intelligence, very different groundedness.

Can it also generate automated tests?

Yes: the same requirement context can drive Playwright or any framework, with the code landing in your repo while the linked test cases land in BesTest. See the AI testing hub for that flow.

Ready to Get Complete
Testing Visibility?

Start using BesTest today and experience complete testing visibility from requirements to release. Free to get started!

Available on the Atlassian Marketplace. Install in seconds, free for up to 10 users.

Start Free on Marketplace
Enterprise SecurityBuilt on Atlassian Forge.Built by TestersDesigned by QA professionals.Free to StartGet started for free and scale as you grow