Alpha

What's new in QA Hub

Every release in the open. We ship fast and we tell you what changed.

v1.4 May 2026 New

Smarter Ticket Discovery with AI

"Ask AI to find tickets" now matches what a senior QA engineer with full tool access would surface. Two-pass ranking, epic-aware expansion, and transparent reasoning.

  • Two-pass ranking — first pass shortlists candidates from titles + labels + epic, second pass re-ranks the shortlist using full acceptance criteria. Closer to how a human reviewer reads.
  • Epic expansion — when your query signals a multi-ticket flow ("E2E", "regression", "suite", "flujo completo"), QA Hub pulls sibling tickets that share an epic so the AI can rank a complete user journey instead of disconnected matches.
  • Reasoning banner — results now include a short explanation of WHY those tickets were picked, in the same language as your query.
  • Better keyword extraction — domain terms (regression, flow, e2e, qa, uat, ios, api) survive the filter; common command words ("give me…", "dame los…") get stripped. No more hardcoded biases.
v1.3 May 2026 New

Autonomous QA Agent, K6 Performance & Plan Approvals

QA Architect goes autonomous. It now reads your tickets from Linear or Jira before writing test plans, picks the right test strategy on its own, generates k6 performance scripts, and routes plans through a QA Manager approval before cases go live.

  • Ticket context — the new load_tickets tool pulls issues from your connected Linear or Jira workspace and feeds them to the agent as context, so generated test cases actually reflect what was built.
  • Automatic strategy selection — select_test_strategy analyzes tickets and picks the right type: Quick Smoke for P0/P1 hotfixes, Performance for load keywords, Security for auth/RBAC scope, API Contract, E2E Full, or Regression.
  • K6 load test generation — when the strategy is Performance, generate_k6_script writes a complete k6 v0.49+ script with staged VU ramps, p95 < 500ms thresholds, and check() assertions. Ready to paste into K6 Studio.
  • Plan approval flow — generated test cases group into an ArchitectPlan with PENDING_REVIEW status. QA Manager gets an in-app notification and approves from a new QA Plans page. Cases move to READY only after approval.
  • Public share link — every plan gets a read-only public URL (/share/<token>) that works without login, so you can share the plan with stakeholders before it goes into the test library.
  • AI token usage — Settings → AI Model now shows a 30-day breakdown of token consumption by operation (generate-plan, generate-bdd, generate-k6, etc.).
  • run_full_cycle — single entry point that chains bootstrap, ticket loading, strategy selection, and codebase analysis in one call. Just point it at a project and let it work.
v1.2 May 2026 New

Architect Report, Agent Mode & Test Quality

QA Architect now generates test plans using the AI model already in your coding agent — no external AI dependency. Includes a dedicated results dashboard, step enrichment from real UI, and reliable result ingestion.

  • Architect Report — dedicated dashboard at /architect showing pass rate by suite, recent failures with TC codes, and a run selector to inspect any specific execution.
  • Agent mode — the Test Architect MCP now lets your AI coding agent (Claude, etc.) generate the ISTQB test plan and Playwright specs directly from the codebase scan, without relying on a server-side AI call.
  • Step enrichment — new enrich_plan tool rewrites vague test steps ("Click Run") with real UI element names captured from a headless browser crawl ("Click the 'Run calculation' button in the toolbar").
  • TC code tagging — generated Playwright specs prefix test titles with TC codes so execution results automatically link back to the correct test case in QA Hub.
  • Multi-role parallel testing — up to 6 Playwright workers, each authenticated as a different user role (admin, viewer, etc.). Sessions never conflict.
  • Test quality rules — ISTQB coverage strategy now enforces at least 40% Negative, Boundary, or State Transition scenarios per plan. No more plans with only Happy Path tests.
Alpha 6 May 2026 New

QA Scout, Spec Provenance & AI Context

The biggest release since the MCP Architect. Adds an embedded AI agent with real access to your test data, origin tracking for every test case, and project-aware AI generation.

  • QA Scout — embedded AI agent. Ask it about failing tests, redundancies, metrics, or compare cases. Stays in QA scope. Logs missing features as a wishlist automatically. Responds in your language.
  • Spec Provenance — every test case now tracks its origin: HUMAN, AI_DRAFT, CODE_DERIVED, or HYBRID. Badge visible in the test case library.
  • Project AI Context — store architecture notes and conventions per project. The AI reads them before generating BDD scenarios and test cases, so it stops making generic assumptions.
  • Feature Requests Wishlist — Scout logs requests for missing features automatically. Accessible at /feature-requests. Managers can triage and track status.
  • QA Scout Settings — enable/disable Scout, cap LLM calls per message (1-5), and toggle animations from Settings → AI Model.
  • Auto-versioning — build number now increments automatically on every deploy. No manual bumping.
v1.1.0 April 2026 Release

Marketing Site, Spanish Locale & Settings Redesign

Public marketing site, full Spanish localization, and a complete redesign of Settings as a horizontal tabbed experience.

  • Marketing site — launched qahub.cloud with pricing, how-it-works diagrams, comparison table, and design partner program.
  • Spanish locale — full Spanish (LATAM) translation across every page and AI-generated content using next-intl.
  • Settings redesign — four horizontal tabs: Integrations, AI Model, TestOps, Developer. API tokens, webhooks, and the MCP quickstart all in one place.
  • Test Architect MCP card — step-by-step setup with copy-ready JSON config for Claude Desktop and Claude Code CLI.
  • Mobile navigation — full overlay drawer with language switcher and consistent animation across all pages.
Alpha 5 Feb – Mar 2026 New

MCP Test Architect, CLI & Anthropic Support

Major push on the AI agentic layer. QA Hub can now be driven by AI coding assistants end-to-end: scan codebase, generate plan, sync to library, run tests, report results.

  • Test Architect MCP — four-tool MCP server for Claude Desktop and Claude Code. Scans codebase, generates Gherkin plan, syncs to QA Hub, runs Playwright.
  • @qa-hub/cli — framework-agnostic uploader. Parse JUnit, Playwright JSON, Cypress, or Jest and push results to /api/v1/ingest. Tag tests with @qa-hub('TC-xxx') to map to existing cases.
  • Anthropic Claude provider — full support as a BYOK AI option alongside Gemini, OpenAI, and Ollama.
  • Gemini model upgrade — gemini-3-flash-preview as default, gemini-2.5-flash as stable option.
  • AI generation cancellation — abort in-flight AI calls when the user dismisses the panel.
Alpha 4 Jan 2026 New

Native Metrics, Versioning, Bulk Ops & Audit Logs

Everything needed to actually run QA operations at scale. From version history on every test case to HMAC-signed webhooks firing on every event.

  • Native metrics dashboard — pass rate, flakiness, p50/p95 duration by framework. Data sourced from CI/CD results ingested via the CLI or REST API.
  • Test case versioning — full version history with diff view and one-click restore for every test case.
  • Bulk operations — select multiple test cases and bulk-move, archive, clone, or retag in one action.
  • Result attachments — screenshots and logs attached to individual test run results.
  • Audit logs — append-only log of every mutation in the tenant. Filterable and CSV-exportable.
  • HMAC-signed webhooks — fire on test_run.completed, test_run_result.failed, bug.resolved, and retest.required.
Alpha 3 Late 2025 New

Public REST API, Bug Tracker & Linear Integration

QA Hub becomes programmable. REST API v1 with OpenAPI spec, scoped API tokens, and a full bug tracking loop powered by Linear.

  • Public REST API v1 — /api/v1 with full OpenAPI spec and Swagger UI. Create and query projects, test cases, runs, and results from any CI/CD tool.
  • API tokens — scoped Bearer tokens (read / write) generated from Settings. Hashed with bcrypt, only shown once.
  • Bug tracker — link failed test results to Linear issues. Automatic background polling detects when bugs move to Done and queues a retest.
  • Retest queue — surface all test cases awaiting retest in one view after bugs are resolved.
  • Azure DevOps connector — WIQL queries and CSV import for teams migrating from ADO test plans.
  • Linear pagination — cursor-based pagination for teams with large Linear workspaces.
Alpha 2 Mid 2025 New

Multi-tenant Auth, BYOK AI & Multi-connector

The architecture foundations: every organization gets full isolation, every AI call uses the team's own keys, and tickets flow in from Jira, Linear, or Azure DevOps.

  • Multi-tenant architecture — JWT-based auth with full tenant isolation. Roles: QA Manager and QA Tester. SUPER_ADMIN with impersonation for platform management.
  • BYOK AI — keys stored encrypted (AES-256-GCM). Supports Gemini, OpenAI, and Ollama. No QA Hub markup on AI usage.
  • ISTQB test case generation — AI generates structured test cases (Happy Path, Negative, Edge Case, Boundary, State Transition) directly from ticket content.
  • BDD generation — Gherkin scenarios in English and Spanish from acceptance criteria.
  • Multi-connector — Jira, Linear, Azure DevOps, and Manual Entry. One active at a time, switchable without data loss.
  • Project modules — organize test cases into projects and named modules.
Alpha 1 Early 2025

First working build

Test case library, manual BDD editor, and basic Jira integration. First production deployment.

  • Test case library — create, edit, search, and export test cases with BDD content.
  • Jira connector — search Jira issues and generate test cases from acceptance criteria.
  • Cloud deployment — self-contained deployment with automatic schema migrations on start.
  • Test runs — create test runs, record pass/fail/skip per test case, track progress.