Generate test cases from Jira tickets
Paste a Jira issue key or link and get step-by-step manual test cases with a single expected result each — including the edge cases and negative paths people forget.
Paste a ticket.
Get a test plan.
Checkpoint reads your Jira or Linear ticket — the comments, epic children, linked issues and specs — and turns it into a focused, tickable plan: what changed, what "done" means, and exactly what to click to prove it.
Jira or Linear key, or paste a link
DevQA
✦ Analyze
PROJ-123QATaskIn Progress
Checkout — let shoppers save a card for next time
Shoppers can tick “Save card” when paying, and saved cards are offered at their next checkout — showing only the last four digits.
5/8tested
2/3AC met
◎ Acceptance criteria
AC1A “Save card” option appears at the payment step
AC2The saved card is offered on the next checkout
AC3Guest checkout still works without saving a card
☑ Test tasks
Confirm the change is live on staging
Save a card during checkout
- Add any item to the cart
- Pay with a new card and tick Save card
- Start a second checkout
✓ The saved card is offered, showing only its last four digits
Jira via Atlassian Rovo MCP
Linear via official MCP
Claude · GPT · Gemini · Grok · local models
Native SwiftUI · Liquid Glass
Stop opening ten tabs.
Everything you need to test a ticket lives in one place — researched for you, organised for you, and ready to tick off.
1
Paste a ticket
A key like PROJ-123 or a Jira / Linear link. Choose Dev or QA mode.
2
AI does the research
It reads the ticket, every comment (later ones win), epic children, parent, linked issues and specs — in parallel — and you watch it happen live.
3
Test with a checklist
Numbered steps, one expected result each, acceptance criteria you can tick as met, and every source it read — saved and organised.
Features
Everything a tester actually needs.
Built for the way you verify work — not another dashboard to maintain.
Deep, focused research
Follows comments, epic children, parent, linked issues and Confluence pages or Linear docs — only as far as they change what needs testing.
Tickable plans
Before-you-start setup, numbered steps with one observable expected result, edge cases, open questions. Grouped per ticket for epics.
Acceptance criteria tracking
Tick criteria as met, see which ones no task covers, and track two metrics: tested and AC met.
Every detail, one panel
Description, inline images, attachments, epic children, comments, work log, history and custom fields — in floating glass tabs.
Folders, search, dashboard
Colour-coded folders, smart folders, sidebar search and filters, pins, archive, tags, due dates, batch sprint import, regression suites — and a dashboard for the week.
Live progress
Streamed thinking, per-ticket fetch status, an elapsed timer and a real progress bar — even on huge epics.
Research log
Go back to every lookup and thought behind a plan, with timestamps and duration.
End-to-end scenarios
Optional: realistic user journeys through neighbouring features, built from related tickets — or your codebase, read-only.
Work the plan, not a spreadsheet
Pass, fail or block each task with what happened, add notes, drag screenshots onto a task as evidence, and turn a failure into a pre-filled bug report. Smoke and P0 lenses cut a big plan down to what matters first.
Export & share
Save any plan as Markdown or a beautiful self-contained HTML page — light/dark, printable, tickable in the browser. Perfect for sign-offs and Jira comments.
Time every task
Each task and scenario has its own timer that adds to the plan's total, next to the time estimate for each task.
A tour that shows you around
A spotlight walks you through every part of a real plan — with a sample plan to explore before you've connected anything.
Jira & Linear, any AI
Same features for both trackers — paste a link and Checkpoint picks the right one. Plans by Claude, GPT, Gemini, Grok, OpenRouter or a local model.
iPhone, iPad & sync
Plan at your desk. Test on the device.
The same plans on Mac, iPhone and iPad — ticks, notes and photos follow you.
A real iPhone and iPad app
Analyze a ticket from your phone, work through the plan on the device you're testing, and attach photos straight from the camera as evidence. Research keeps going when you leave the app, with progress on your Lock Screen.
Sync through storage you already use
Pick a folder in iCloud Drive — or your company's OneDrive, Google Drive or Dropbox — and every device stays in step. Each plan is its own file, and the newest edit wins. Your API keys never leave each device's Keychain.
Use cases
One app for QA testers, developers and team leads.
Checkpoint is an AI test plan generator for Jira and Linear: it writes test cases from your tickets with Claude, GPT, Gemini, Grok or a local model, so you can spend your time testing, not reading.
Turn acceptance criteria into a checklist
Acceptance criteria are quoted from the ticket (or derived when missing), mapped to the test cases that cover them, and tracked as met or not met.
Epic & sprint regression testing
Analyze an epic and every active child is covered, grouped per ticket — ready for a release regression pass.
UAT on the hosted app
QA mode writes black-box, UI-only test plans for staging or production — starting with a check that the fix is actually deployed.
Developer self-testing
Dev mode adds branches, APIs, logs and tests, so you can verify your own change before handing it to QA.
Watch it work
Big epic? You'll see every step.
Checkpoint streams the model's reasoning as it happens, fetches tickets in parallel, and shows exactly where it's up to — then keeps the whole log with your plan.
- Parallel fetches. All of an epic's children load at once, each flipping to ✓ as it returns.
- Live thinking. Expand any thought to see why it looked where it did.
- Honest progress. A timer and a bar that fills while reading and writing — not a spinner that lies.
Reading 8 of 8Fetching tickets and docs in parallel
0:42
🎫 9 tickets read⤓ 11 lookups🧠 3 thoughts
This is an epic — fetching its active children and skipping superseded ones, in parallel.
Opened ticket PROJ-123
Opened ticket PROJ-118
Opened ticket PROJ-139
Opened ticket PROJ-140
Read Confluence page
Opened ticket PROJ-142
Writing your test plan · 6.2K characters
Two modes
Written for whoever's testing.
Flip one toggle (⌘⇧M) and the same ticket becomes a very different plan.
Dev
For the developer verifying their change
Technical detail welcome — branches, commits, APIs, logs and tests.
- Which build, branch or commit to verify on
- API responses, logs and data checks where the ticket calls for them
- Unit and integration tests mentioned by the dev
Example stepConfirm PaymentSheet requests saved cards and never receives the full card number.
QA
For a tester on the hosted app
Black-box and UI-only: no code, repos, terminals or local setup.
- First confirms the fix is actually deployed
- Screens, buttons and messages named as you see them
- Roles, empty states, bad input and regressions — technical-only checks become questions for the dev
Example stepAdd an item, go to checkout, and check your saved card appears with only its last four digits.
Your AI. Your key. Or your machine.
Checkpoint works with the model you already trust — hosted or running locally — and lists what your provider offers with one click.
Anthropic ClaudeAdaptive thinking · schema-enforced plans
OpenAIGPT models with tool calling
Google GeminiGemini API key
xAI GrokxAI API key
OpenRouterHundreds of models, one key
Local modelsOllama · LM Studio · vLLM · llama.cpp
Any compatible APIOpenAI- or Anthropic-compatible endpoints
Your tickets stay yours.
Checkpoint runs on your Mac, iPhone or iPad and talks directly to Atlassian, Linear and the AI provider you choose — no Checkpoint servers in between. Run a local model and your tickets never leave your network.
Read-only by design
Jira: only read tools are exposed. Linear: its read-only endpoint and scope, enforced by Linear.
Keys in your Keychain
Your AI provider keys and tracker tokens stay in each device's Keychain — they never sync.
Sign in with OAuth
Standard Atlassian and Linear sign-in with PKCE — no admin allowlisting needed.
Sandboxed & local
Native, sandboxed apps. Plans live on your devices and only sync through a folder you choose.
Up and running in two minutes.
Download
Grab the latest DMG and drag Checkpoint to Applications (macOS 26), or install the iPhone and iPad .ipa from the same release.
Follow the welcome
First launch walks you through Dev or QA, your AI (Claude, OpenAI, Gemini, Grok, OpenRouter or a local model, with a live key test), your trackers and sync — then offers a quick tour.
Connect your tracker
Sign in with Atlassian and/or Linear, hit Test connection, then paste your first ticket.
Questions, answered.
How do I generate test cases from a Jira ticket?
Install Checkpoint, choose an AI provider, sign in with Atlassian, then paste the ticket key (for example PROJ-123) or its link. Checkpoint reads the ticket, its comments, linked issues, epic children and Confluence specs, and writes a numbered test plan with expected results in about a minute.
Does Checkpoint work with Linear?
Yes. Sign in with Linear or use a Linear API key. Linear issues get the same test plans, Dev and QA modes, ticket panel, folders and acceptance-criteria tracking as Jira.
Can it turn acceptance criteria into test cases?
Yes. Explicit acceptance criteria are quoted from the ticket, every criterion is mapped to the test cases that cover it, and criteria with no covering test are flagged so nothing slips through. You can tick each criterion as met.
Which AI model does Checkpoint use?
Whichever you like: Anthropic Claude, OpenAI, Google Gemini, xAI Grok, anything on OpenRouter, or a local model. Bring your own key, pick the model from a live list, and set how much effort it spends — lower for quick plans, higher for large epics. Claude gets extras like adaptive thinking and schema-enforced plans.
How is this different from a test case management tool?
Test management tools store test cases you write by hand. Checkpoint writes the test plan for you from the ticket itself, keeps it next to the full ticket details, and tracks what you've tested — no spreadsheet or separate tool to maintain.
Does it work with OpenAI, Gemini or Grok?
Yes. Checkpoint supports OpenAI, Google Gemini, xAI Grok and OpenRouter alongside Anthropic Claude. Paste your API key, fetch the model list, and pick any model that supports tool calling.
Can I use a local LLM like Ollama or LM Studio?
Yes. Point Checkpoint at any OpenAI-compatible server (Ollama, LM Studio, vLLM, llama.cpp) or Anthropic-compatible endpoint (such as LiteLLM) on your machine or network. No API key needed, and ticket content stays local. Use a model with tool calling and a large context window for big epics.
What are test scenarios?
An optional extra: besides the step-by-step test cases, Checkpoint can write 3–6 end-to-end user journeys that exercise the change in context — through the screens and features around it — with a role, goal, steps and end result. Build them from related tickets alone, or also let Checkpoint read your codebase (read-only, only the folder you choose) for journeys that match how the product really works.
Can I build Checkpoint from source?
Yes. Clone the GitHub repository and run ./build.sh — it checks for Xcode 26 and XcodeGen and produces a universal Checkpoint.app in the dist folder. Add --install to copy it to Applications.
Can I share or export a test plan?
Yes. Copy a plan as Markdown for Jira comments and pull requests, or save it as a self-contained HTML page styled like the app — light and dark mode, print-friendly, with checkboxes anyone can tick in their browser.
Is Checkpoint free?
Yes — it's free and open source under the MIT license. You bring your own AI provider key and pay that provider directly for what your plans use — or run a local model for free.
Can it change my tickets?
No. Checkpoint only ever reads. For Jira it exposes read tools only; for Linear it uses Linear's read-only MCP endpoint with a read-only scope, so Linear itself blocks writes.
Does my data go through your servers?
There are no Checkpoint servers. The app talks directly to your tracker's MCP server and to the AI provider you choose. Plans are saved on your devices, and only sync through a folder you pick (iCloud Drive, OneDrive, Google Drive or Dropbox).
Which devices does it run on?
Any Mac on macOS 26 or later (a universal app for Apple Silicon and Intel), and iPhone or iPad on iOS / iPadOS 26 or later.
Is there an iPhone and iPad app?
Yes. The iOS app has the same plans and features: analyze a ticket on your phone, work through it on a test device, attach photos from the camera as evidence, and keep research running when you leave the app. Each release includes an .ipa you can install with AltStore or Sideloadly, or build it yourself in Xcode.
How do my plans sync between Mac and iPhone?
Choose a sync folder in iCloud Drive, OneDrive, Google Drive or Dropbox on each device. Checkpoint keeps each plan as its own file there, merges changes by the most recent edit, and syncs ticks, notes, tags and evidence. On a work machine, point it at your company OneDrive or Google Drive so nothing goes to a personal iCloud. API keys and sign-ins never sync.
My org restricts Atlassian API tokens. Will it work?
Use Sign in with Atlassian. It uses OAuth with a local redirect that Atlassian pre-approves for every org, so no admin change is usually needed.
macOS says the app can't be opened.
If a release isn't notarized yet, right-click Checkpoint in Applications and choose Open the first time, or run xattr -dr com.apple.quarantine /Applications/Checkpoint.app.
Test the ticket,
not your patience.
Download Checkpoint for Mac, iPhone or iPad and turn your next ticket into a plan in about a minute.