Borsch documentation
How to install, configure, and get the most out of Borsch on Jira and Azure DevOps.
How it works
Borsch appears where your work already lives — an issue panel in Jira Cloud and a work-item tab in Azure DevOps. It reads the item's title, description, acceptance criteria, and human comments, then runs a multi-stage AI pipeline to produce a refined result.
Stage 1 — Draft
A first AI pass generates the initial output from the full context — test cases across positive, negative, and edge scenarios, a structured bug report, or a complete user story.
Stage 2 — Review & refine
A second AI pass reviews the draft against the original requirement: fills coverage gaps, removes duplicates, sharpens wording, and scores overall quality 1–10.
Stage 3 — Optimize (test sets)
On Azure DevOps, sets of 3+ test cases get a final pass that detects shared preconditions and overlapping coverage. It lifts shared setup into preconditions and merges cases that substantially duplicate each other, and shows you what it grouped. This pass does not run on Jira, where generation stays inside Atlassian's execution limit.
A quality score and the gaps it found are shown after each run, so you can see how good the output is before you use it.
What Borsch does
Test cases
Turn a user story into automation-ready test cases with categories, priorities, preconditions, numbered steps with expected results, and test data. Includes Sprint Coverage Radar and an Incident-to-Test bridge.
Bug reports
Turn a rough description into Steps to Reproduce, Expected Result, Actual Result, and a justified Severity — with missing-information detection and duplicate detection.
User stories
Enrich a bare title into a full "As a… I want… so that…" story with acceptance criteria and a Definition of Done, or break an epic into implementable stories.
Generating from a work item
- Open any work item — a Story, Task, or Bug in Jira, or a User Story / Bug in Azure DevOps.
- Open the Borsch panel.
- Pick what you want: test cases, an enhanced bug report, or an enriched story.
- Optionally choose an output format, a strategy template, and the maximum count.
- Click Generate and wait for the multi-stage pipeline to finish.
- Review the results, select what to keep, and save them as work items (or sub-tasks on Jira).
Borsch reads items it has already created on the same work item and avoids generating duplicates.
Output formats
Test Cases (default)
Structured test cases with title, category (positive / negative / edge), priority, preconditions, numbered steps with expected results, test data, and optional automation hints.
BDD / Gherkin
Given / When / Then scenarios, with Scenario Outline and Examples tables for parameterized tests.
Code (Playwright / Cypress / Selenium)
Copy-paste-ready test code for your chosen framework, using framework-specific syntax, locators, and assertions.
Strategy templates
Strategy templates guide the AI on what kind of tests to generate and pre-set sensible defaults.
| Strategy | Focus |
|---|---|
| Comprehensive | Full coverage — positive, negative, edge |
| Smoke | Critical happy-path only |
| Regression | Areas likely to break from code changes |
| Security | OWASP Top 10 — auth, injection, access control |
| Performance | Load, concurrency, timeouts, boundary values |
| Accessibility | Keyboard navigation, screen readers, ARIA, contrast |
Project Knowledge
Project Knowledge teaches Borsch about your tech stack, UI components, business rules, and testing conventions. With it, output references your actual technology and follows your project's naming patterns instead of reading generic.
To configure, open Borsch's settings for the project and fill in any combination of: Tech Stack, UI Components, Business Rules, and Testing Conventions. All fields are optional — Borsch works without Project Knowledge, it simply produces more generic output.
Wiki context (Azure DevOps)
On Azure DevOps you can link Wiki pages as additional AI context — useful for requirements docs, API specs, or domain glossaries. Borsch reads the page text and includes it when generating.
Wiki context is optional. Borsch generates from the item's description and acceptance criteria on its own; the wiki step only adds extra context when you have pages worth including. Borsch reads text content only — attachments and embedded images are not processed, and large pages are truncated to fit the AI context.
On Jira, Borsch does not read Confluence. It requests Jira permissions only, and takes its context from the issue itself plus your Project Knowledge.
Xray & Zephyr Scale integration (Jira)
This one is Jira only. By default Borsch saves test cases as Jira sub-tasks, and you can configure it to create tests directly in Xray or Zephyr Scale instead. On Azure DevOps, test cases are always saved as Test Case work items.
Add your Xray API credentials (Client ID + Client Secret) or a Zephyr Scale API token in Borsch's TMS Integration settings and test the connection. If credentials are missing or a connection fails, Borsch automatically falls back to creating work items so no test cases are lost.
Plans and billing
Borsch is billed by the marketplace you installed it from, on that marketplace's own model. Exact, current prices are published there rather than repeated here.
| Platform | How it is billed | Where to buy |
|---|---|---|
| Jira Cloud | Per user, by Atlassian, on your existing account | Atlassian Marketplace |
| Azure DevOps | Per organization, from a monthly credit pool | Visual Studio Marketplace |
On Azure DevOps, one credit is one AI generation, and the pool is shared across test cases, bug reports and user stories for the whole organization however many people use it. The count resets on the first day of each calendar month. On Jira the allowance comes with the Atlassian licence instead, so it does not depend on a plan chosen here. Both platforms have a free tier.
FAQ
Which platforms does Borsch support?
Jira Cloud via the Atlassian Marketplace and Azure DevOps via the Marketplace. The same AI pipeline runs on both, adapted to each platform's work items.
Where is my content sent?
The work item's title, description, and acceptance criteria are sent to the Borsch API for AI processing and are not stored after the result is returned. See our Privacy Policy.
Can my whole team share one quota?
Yes. The credit pool is shared across all users in the same organization or project. There is no per-seat limit.
Why does a run take a little longer than single-pass tools?
Borsch runs multiple sequential AI passes — draft, then review, then optimize for test sets. Each pass is a separate call. The extra time is what produces higher-quality output than generate-once tools.
Do I need Project Knowledge or wiki pages?
No — both are optional. Borsch works from the item's description and acceptance criteria alone; adding context simply tailors the output to your project.
Support
Questions not covered here? Contact us at [email protected] or visit our support page.