Your AI test engineer, from ticket to pull request.
Paste a Jira ticket, a pull request or a failed CI run. TestTale writes end-to-end tests in your framework, runs them, heals what broke and flags real bugs — then waits for your review before opening the PR.
No card needed · Bring your own AI key · Cancel any time
Writes idiomatic tests for
- Playwright · TypeScript
- Cypress · TypeScript
- WebdriverIO · TypeScript
- Playwright · Python
- Selenium · Java
How it works
Start from what you already have
One prompt box. The agent figures out what you pasted and gathers the right context.
Jira ticket
Paste a key or URL. Acceptance criteria are extracted — from a custom field, a heading, or Gherkin.
Pull request
Tests for exactly what the diff changed, plus the linked ticket's acceptance criteria.
Failed CI run
Every failure is classified — test bug, product bug, flaky or environment — with a verified fix.
Plain requirement
No ticket yet? Describe the behaviour and get a reviewable suite.
- 1
Gather
Ticket, ACs, diff or CI logs — plus your repo's conventions.
- 2
Generate
Scenarios mapped to every AC, written in your framework.
- 3
Execute & heal
Runs in a sandbox, repairs test defects, flags product bugs.
- 4
Review & ship
Edit or ask the agent, then open the PR when you're happy.
Features
Tests your team would actually merge
Built for QA engineers and developers who care about coverage, readability and trust.
Mapped to acceptance criteria
Every scenario cites the ACs it covers, so gaps are visible before review — and in the PR.
Your framework, your conventions
Reads your config, page objects and fixtures first, then writes code that looks like your team wrote it.
Runs and self-heals
Executes every suite in an isolated sandbox and repairs broken locators and waits from real failure evidence.
Finds real bugs
When the app is wrong, the test stays strict and the bug is flagged — with evidence — instead of being “healed” away.
Human in the loop
Edit in the browser or ask the agent in plain English. Every change is versioned with a diff and one-click undo.
Pull-request ready
Branch, commit and a PR description with a coverage matrix — created only when you click the button.
The QA workspace
Everything your QA practice needs, around the tests
Writing tests is where it starts. Plan, review, schedule and ask questions in the same workspace.
QA assistant
Ask in plain words: what failed last night, which tests are flaky, what a ticket still needs. Anything it wants to change waits for your approval.
Test library and plans
Manual and automated test cases in one place. Group them into plans, run them by hand, and file a Jira bug from any failure.
Story reviews
Before work starts, get a testability score, open questions, missing cases and suggested acceptance criteria for any story.
Regression and flaky tests
Run your existing suite every hour, day or week. See the pass-rate trend, the tests that flip between pass and fail, and the slowest ones.
One place to start the day
Home shows what needs you: runs to review, approvals, failing schedules, flaky tests and stories that need answers.
Connects to your tools
GitHub and Jira in one click, plus any MCP server, so the agent can read the context your team already keeps elsewhere.
Bring your own AI key
Your model, your account, your costs. No markup.
Add a key from the AI provider your company already uses, or any compatible endpoint, including self-hosted models. TestTale sends your tickets and code only to that model, under your account, and your provider bills you directly at its own prices.
Typically cents per run
A test-generation run usually costs from a few cents to about a dollar of model usage, depending on the model you pick and how much code it reads.
Pick per task
Set a default model, a faster one for small jobs, and switch for any single run or conversation.
Try it without a key
A demo agent shows the whole flow on sample data until you add one.
Self-healing, done honestly
Fix the test when the test is wrong. Report the bug when the app is.
After every run, each failure is analysed with its error, stack trace and a snapshot of the page. The agent decides what the failure means — and only touches code when the test itself is at fault.
- FixStale locators, missing waits, wrong navigation or test data.
- FlakyTiming issues are hardened without weakening assertions.
- Product bugThe app violates an acceptance criterion — the test stays strict.
- EnvironmentApp down, missing credentials or network errors — no code changes.
Attempt 1
3 / 5 passed
1 locator fixed · 1 product bug flagged
Attempt 2
4 / 5 passed
Remaining failure is the product bug — not healed
Review
Ready
PR description calls out the bug with evidence
Security
Built to be trusted with your codebase
Guardrails are part of the architecture, not a policy document.
Approval for every change
The AI can read, but not change. Pull requests, Jira bugs and comments, and other write actions happen only after a person approves them.
Tests run in your CI
Generated tests run in your own GitHub Actions, through a workflow you add with one reviewed pull request. Your code stays in your CI.
Encrypted secrets
AI keys, integration tokens and test credentials are encrypted at rest and redacted from logs.
Least-privilege access
GitHub App and Jira OAuth scopes are limited to what the product needs, and you can revoke them at any time.
Audit trail
Who generated, edited, healed and approved what — recorded per workspace.
Tenant isolation
Every record is scoped to your workspace, with role-based access for owners, admins and members.
More detail on our security page.
Pricing
Simple per-user pricing
Start with a 14-day free trial of every feature, no card needed. Runs are unlimited; you bring your own AI keys, so model usage is billed by your provider at cost.
Team
$8 / user / month
Billed yearly, or $10 billed monthly
- Unlimited runs (fair use)
- All 5 frameworks, self-healing, review and PRs
- Bring your own AI provider and key
- 3 parallel runs · 10 projects · 3 MCP servers
- Daily regression schedules
- 30-day audit log
Business
Most popular$16 / user / month
Billed yearly, or $20 billed monthly
- Everything in Team
- 10 parallel runs
- Unlimited projects and MCP servers
- Hourly regression schedules
- 1-year audit log
- Priority support
Enterprise
Custom
For larger teams with security reviews
- Everything in Business
- SSO / SAML
- Custom data retention
- Invoicing and procurement
- Dedicated onboarding
Prices in USD, excluding taxes. Payments are handled securely by our reseller Paddle.com. New subscriptions have a 14-day money-back guarantee.
FAQ
Questions, answered
Which tools does it work with?+
Jira Cloud for tickets, GitHub for code, pull requests and Actions, any MCP server for extra context, and five test stacks: Playwright (TypeScript and Python), Cypress, WebdriverIO and Selenium (Java). You pick the framework per run.
Do I need an AI key to try it?+
No. Without a key, a demo agent shows how everything works on sample data. Add your own key when you want it to work on your tickets and code.
Where do the tests run?+
In your own GitHub Actions. One reviewed pull request adds the workflow to your repository; after that, every run uses your CI and your test environment.
Does it need access to my code?+
Read access to the repositories you choose, so it can follow your existing conventions and page objects. Write access is used only to create the branch and pull request after you approve it.
What exactly does “self-healing” change?+
Only test code that is provably wrong — a stale locator, a missing wait, wrong navigation or test data. Each fix is a new file version with a diff and an explanation you can undo.
Can self-healing hide a real bug?+
It's designed not to. Assertions that verify an acceptance criterion are never loosened. If the application disagrees with the ticket, the failure is reported as a product bug, and you can file it in Jira in one click.
Which AI model does it use?+
The one you choose. Connect your own AI provider with your own key, and your provider bills you directly. Your tickets and code are sent to that model only to generate, heal and triage tests for your workspace.
How much does it cost?+
Team is $8 per user per month billed yearly ($10 monthly); Business is $16 ($20 monthly). Every workspace starts with a 14-day free trial, no card needed. Runs are unlimited, and you bring your own AI keys, so your model provider bills you directly for model usage.
What happens when the trial ends?+
Your workspace becomes read-only: nothing is deleted, and you can still view and export everything. Choose a plan to keep working.
Can I cancel any time?+
Yes, in Settings → Billing & usage. You keep access until the end of the period you've paid for. New subscriptions have a 14-day money-back guarantee.
Start today
Get your first AI-written, self-healing suite this week.
Create a workspace, connect GitHub and Jira, and paste your first ticket. Every feature is included for 14 days, free.
Questions first? Email us and we'll help you set up.