Tracing, evaluation, red-teaming, human review and exportable evidence across the AI lifecycle, next to 35+ testing modules for the software around it. Every AI call uses your own model keys.
Free plan: $0/month for 3 members with capped AI evaluation — no credit card required. Compare plans
For the engineers who ship AI features and the people who sign them off
Step 1 of 8 · Smoke-test a model
Model Playground
FreeSend a prompt to a model on your own key and read tokens, latency and estimated cost.
Connect your provider key to run a model. Provider usage is billed separately.
Conversation
System
You are the support assistant for Example Co. Answer only from the policy text.
User
Can I get a refund 45 days after buying?
Refunds are available within 30 days of purchase. After that, the policy offers account credit instead of a refund (section 4.2).
Illustrative sample data, modelled on the app’s screens.
For the engineers who build AI features and the people who approve them. Every step uses your own model keys and lives in the same workspace as your software tests.
Send LLM and agent spans over OpenTelemetry (OTLP/HTTP) and read conversations turn by turn. Free keeps 10,000 spans a month for 7 days; Pro 2,000,000 for 90 days.
Score against versioned datasets, public benchmark samples and custom scorers. On Pro, evaluate live spans and whole conversations, watch for drift and test prompt rewrites before you accept one.
Probe models and agents with adversarial prompts before release. Free runs 10 probes per run; Pro runs the full red team.
Route spans and results to human review queues, measure reviewer agreement, and turn labels into ground truth.
Gate releases in CI and export evidence records, audit logs and AI governance reports for the people who sign off (Pro).
See what each step includes on Free and Pro. New to evaluating AI features? Read the LLM evaluation guide, or see how RAG, agent and red-team evaluation work.
Compliance evidence mapping — not a certification
SOC 2 evidence mapping
PCI-DSS evidence mapping
GDPR evidence mapping
HIPAA evidence mapping
NIST 800-53 control mapping
FISMA evidence mapping
Free tool · no signup
Audit any website’s redirects, security headers, CSP, caching and SEO in about 20 seconds — with suggested fixes.
Testing Modules
Trace spans / month on Pro
Frameworks with evidence mapping
Supported AI Providers
AI evaluation features plus 35+ testing modules organized by discipline, with autonomous agents on Pro. Every AI feature runs on your own provider key.
Director + 6 Worker Agents
Provide a URL, API spec, LLM endpoint, or AI agent config — and our director orchestrates web, API, LLM, agent, research, and report agents to autonomously plan, execute, and summarize quality tests. A ReAct loop engine, a rule that every finding must cite tool evidence, and per-domain memory are included.
Agents crawl and discover what to test
OWASP Top 10 with real evidence
WCAG 2.2 & Section 508 compliance
Core Web Vitals & load testing
30+ adversarial attack patterns
Health score with remediation steps
OpenTelemetry (OTLP/HTTP) ingest; SDK setup for LangChain, LlamaIndex, OpenAI Agents, OpenAI and Anthropic
Group traces by session and read each conversation turn by turn
Versioned datasets, custom scorers and checks on agent runs
Score sampled live spans or whole conversations and alert when scores drop
Score PSI, pass-rate drop and opt-in embedding drift against a baseline window
Adversarial probes; full red team on Pro
Route spans and results to reviewers, measure agreement
Immutable versions, staging and production labels, diff, rollback and a playground
Candidates from failing cases, tested in one experiment; nothing changes until you accept one
Versioned datasets, public benchmark samples, scorers incl. WER/CER, JavaScript code scorers (Pro), MCP checks, calibration, SDKs and CLI
Faithfulness, answer and context relevance, judged claim by claim by an independent model
Attack probes against a model or agent, graded per probe and mapped to the OWASP LLM Top 10
Scenarios sent to your own HTTPS agent endpoint, with tool-call and budget checks
Image, audio and PDF cases sent to your vision or audio model and scored
Compare variants with a significance test; gate releases in GitLab, Jenkins, Azure DevOps or GitHub Actions with JUnit reports
Hash-chained evaluation records, evidence packs and AI governance exports
Real-time QA metrics, test execution trends, and insights
Hierarchical suites, drag-and-drop organization, RBAC
Click on a live browser stream — AI generates full suites
Automated test scheduling and execution orchestration
REST and GraphQL with collections, environments, chaining
Live schema, read-only queries, EXPLAIN plans
Lab page metrics: LCP, CLS, FCP, TBT, TTFB
Concurrent virtual users, p50/p90/p95/p99 latency
7 breakpoints, device emulation, network throttling
Connect, send/receive, stress test, protocol validation
Query/mutation execution, schema introspection
Pact-based consumer-driven contract testing
OWASP-aligned vulnerability and threat detection
15+ vulnerability rules, 10 secret patterns, SCA
Section 508 and WCAG 2.0/2.1 rules with axe-core
GDPR, HIPAA, PCI-DSS, SOC 2 multi-framework
76 NIST 800-53 moderate baseline controls
Completeness, consistency, duplicates, outliers
Latency injection, error injection, recovery testing
Smart contract verification and validation
Pixel-perfect UI comparison and visual diff
Multi-language and locale testing automation
OpenAI, Anthropic, Gemini, DeepSeek, Mistral, Groq, Azure OpenAI
Hallucination, bias, robustness, safety testing
Red team, drift, RAG, multi-turn, rubrics, explainability
Plain English in — executable tests out
AI auto-fixes broken selectors when UI changes
Module coverage metrics and gap analysis
GitHub Actions runner and GitHub App checks; GitLab CI, Jenkins, Azure DevOps and CircleCI config templates
Application performance monitoring & analytics
HTML, JSON and PDF exports with executive summaries
Your S3, R2 or S3-compatible bucket, chosen per data type
Your own OIDC or SAML provider, DNS-verified domains, required SSO
Admin, tester and viewer roles plus organization owner, admin and member
TOTP MFA, account lockout, password policy
CSV or JSON Lines export; signed HTTPS webhooks to your SIEM
Native observability and metrics export
REST endpoints you can call with an organization API key
A short, checkable case. Try it on Free before you decide.
Tracing, evaluation, red-teaming and human review sit next to API, security, accessibility and performance testing, so one release review covers the model and the software around it.
Every AI feature uses your own provider key. Your provider bills you directly, and platform credits never stand in for model tokens.
On Pro, export evidence records, audit logs (CSV or JSON Lines) and your organization's data, and stream the audit log to your SIEM.
Pro is $199 per workspace per month with 10 members included and $29 per extra member. No automatic overages.
Every AI feature on Free and Pro requires your own provider key. GitHub Actions workflows use secrets in your repository. Your provider bills API usage separately from platform execution credits.
NexGen QA dispatches tests to your GitHub Actions environment. The runner authenticates with your AI provider using secrets you control, executes the test suite, and posts results back to NexGen over an authenticated callback for live status display. AI provider keys stay in your GitHub Secrets and tests run in your own runner.
Settings → BYOK → register a secret name per provider (e.g. OPENAI_API_KEY).
Repo Settings → Secrets and variables → Actions → New secret.
Paste a contents:write + actions:write token, repo owner, repo name.
NexGen commits .github/workflows/nexgen-qa.yml and tests/nexgen/sample.test.ts.
Pick suite + provider → Dispatch. Live results stream back to NexGen.
Register any OpenAI-, Anthropic-, or Gemini-compatible endpoint — or use raw passthrough for any HTTP API. Your custom models appear in every AI testing model picker across the platform.
Tests execute in your repo’s Actions and report back over an authenticated callback. Pin the runner version, audit every dispatch in your repo’s Actions history, and integrate with your existing branch protection rules.
AI calls require your own provider credentials. GitHub Actions workflows read repository secrets.
The runner posts results back with an organization API key or a per-run token.
Tests execute in your repository's Actions runners, GitHub-hosted or self-hosted.
Jira issue creation, GitHub Actions and GitHub App checks, CI config templates, Slack and webhook alerts, SSO with your identity provider, and OpenTelemetry and Prometheus for observability.
Need another tool? Outbound webhooks connect NexGen QA test events to any system that accepts HTTP callbacks.
How teams in regulated and high-traffic industries use NexGen QA's modules, from healthcare data handling to federal accessibility requirements.
HIPAA- and GDPR-focused compliance testing with automated PHI detection. API, security and accessibility testing for clinical systems and EHR platforms.
Test automation for trading platforms and core banking systems. Security, API and load testing help catch transaction-path defects before each release.
Visual regression across breakpoints and browsers, plus load testing before peak traffic events.
Section 508 and WCAG 2.0/2.1 scanning, NIST 800-53 control mapping with POA&M tracking, and Login.gov sign-in. NexGen QA holds no FedRAMP or DoD authorization; on-premise deployment is by separate agreement.
Free covers everyday QA and capped AI evaluation. Pro adds online evals, release gates, the full red team and enterprise controls for a flat workspace price. Every AI feature uses your own provider key; provider charges are separate.
Everyday QA, deterministic website checks and capped AI evaluation with your own model keys. No credit card required.
For teams shipping AI features. $1,990 per workspace when billed annually. 10 members included; extra members $29/month or $290/year. AI provider charges are separate.
New prices apply to new subscriptions. Existing subscribers keep their current price, and we will give at least 30 days' notice before any change. Need a DPA, invoicing or on-prem? Talk to us.
Pro subscriptions renew at the selected monthly or annual interval until you cancel. Cancel anytime from Manage Your Subscription or by emailing contact@qa-automation.com; access continues through the paid period. All sales are final except where required by law. See the Terms of Service and Privacy Policy.
Already a subscriber?
Upgrade, downgrade, update payment methods, or cancel your plan. Stripe will email you a one-time code to confirm it's you.
TLS in transit · No automatic overages · contact@qa-automation.com
Identity, storage, retention, model choice and audit delivery are yours to set. Here is exactly what each control does, and what NexGen keeps running.
Bring your own OIDC or SAML identity provider, on domains you verify by DNS, and require SSO for your whole organization with a break-glass owner.
Keep generated reports, datasets, multimodal artifacts, visual-regression screenshots, evidence snapshots and exports in your own S3, R2 or S3-compatible bucket, chosen per type, for data written after you enable it.
Set retention per data type. Export your organization's data as a ZIP with an expiring link. Delete it with a verified, signed receipt.
Choose which AI providers and models may run, set a monthly spend limit (estimated), and choose whether evaluation and trace prompts and responses are stored in full, redacted, or as metadata only.
Stream your audit log to your SIEM over signed HTTPS webhooks, with at-least-once delivery.
NexGen hosts the application, its database and orchestration. The controls above decide what is stored, where your bucket-backed files live, and who can sign in; they do not move the service into your cloud.
Walk through tracing, evaluation, red-teaming and human review on your own use case, alongside the testing modules you need. Email contact@qa-automation.com or schedule a demo below.
These are arranged by separate agreement and are not part of the published plans. NexGen hosts the application, its database and orchestration.
Talk to us: contact@qa-automation.com
Straight answers on BYOK, custom AI models, the 35+ testing modules, security, compliance, deployment, and pricing — written for engineers, buyers, and AI assistants alike.
NexGen QA puts tracing, evaluation, red-teaming, human review and exportable evidence for AI features in the same workspace as its software testing modules. You send LLM and agent traces over OpenTelemetry (OTLP/HTTP), with SDK setup for LangChain, LlamaIndex, the OpenAI Agents SDK, OpenAI and Anthropic, and read conversations turn by turn. You evaluate against versioned datasets and samples of public benchmarks, run red-team probes, route outputs to human review queues, manage prompt versions and gate releases in CI. Pro adds online evals for live spans and whole conversations, drift monitors, a prompt optimizer whose candidates a person accepts, and sandboxed JavaScript code scorers. Every AI call uses your own model keys.
No tool does, and NexGen QA does not claim to. Volume and retention are capped by plan: Free keeps 10,000 trace spans a month for 7 days, Pro 2,000,000 a month for 90 days. Customer storage applies to data written after you enable it. NexGen QA does not host or serve your models. Check the features you depend on against a free workspace before you buy.
Bring your own OIDC or SAML identity provider, on domains you verify by DNS, and require SSO for your whole organization with a break-glass owner. Keep generated reports, datasets, multimodal artifacts, visual-regression screenshots, evidence snapshots and exports in your own S3, R2 or S3-compatible bucket, chosen per type, for data written after you enable it. Set retention per data type. Export your organization's data as a ZIP with an expiring link. Delete it with a verified, signed receipt. Choose which AI providers and models may run, set a monthly spend limit (estimated), and choose whether evaluation and trace prompts and responses are stored in full, redacted, or as metadata only. Stream your audit log to your SIEM over signed HTTPS webhooks, with at-least-once delivery. NexGen hosts the application, its database and orchestration. The controls above decide what is stored, where your bucket-backed files live, and who can sign in; they do not move the service into your cloud.
AI test automation uses AI models to generate, run and maintain software tests instead of only executing hand-written scripts. In NexGen QA that means describing behavior in plain English to generate executable test scripts with selectors, assertions and test data, then letting self-healing tests repair broken selectors automatically as the UI changes, across 35+ testing modules covering API, security, performance, accessibility, database and more.
Yes. You write test cases in plain English and the AI engine — models from OpenAI, Anthropic, Google (Gemini), DeepSeek, Mistral, Groq and Azure OpenAI, or a custom model you register — converts the description into executable test scripts. Self-healing tests then detect when an element selector breaks because the UI changed and repair it automatically, which is aimed at the recurring cost of test maintenance rather than just the one-time cost of writing tests.
It depends on the mode. With the BYOK GitHub Actions runner, tests execute inside your own GitHub Actions runner using AI provider keys stored in your own GitHub Secrets. Otherwise NexGen hosts the application, its database and orchestration, and retains account and access metadata, test definitions and results; in-app provider keys are encrypted at rest. On Pro you can keep generated reports, datasets and other files in your own bucket for data written after you enable it, and store prompts and responses redacted or as metadata only.
BYOK stands for Bring Your Own Key. Every AI feature on Free and Pro requires a customer AI provider key. You can register AI provider keys — for OpenAI, Anthropic, Google (Gemini), DeepSeek, Mistral, Groq and Azure OpenAI — as GitHub Secrets, or register any OpenAI-, Anthropic-, or Gemini-compatible endpoint (including a self-hosted model) from Settings → Custom Models. Custom models then appear in every AI testing model picker across the platform.
Register a secret name in NexGen QA, paste a fine-grained GitHub personal access token, and NexGen QA commits a workflow file and a sample test to your repository. When you dispatch a run from the CI/CD tab, tests execute inside your own GitHub Actions runner using the AI provider key stored in your own GitHub Secrets, and results are posted back to the platform over an authenticated callback.
NexGen QA covers 35+ testing modules: API testing (REST and GraphQL), security testing (DAST, SAST, OWASP Top 10), accessibility testing (WCAG 2.0/2.1 rules and Section 508), performance and load/stress testing, mobile and responsive testing, database testing, WebSocket testing, contract/Pact testing, visual regression testing, chaos and resilience testing, blockchain testing, localization/i18n testing, and AI model validation — plus compliance auditing and CI/CD integration.
Yes. The AI Validation Suite covers hallucination detection, bias detection, robustness testing under adversarial inputs, toxicity scanning and safety evaluation. Advanced AI Testing adds red teaming against 30+ adversarial attacks, model drift detection, RAG validation for retrieval accuracy and grounding, and prompt-injection resistance testing.
It depends on what you need. Katalon is a mature, scripted test automation tool spanning web, mobile, API and desktop testing with a free tier and test management. mabl focuses on agentic, low-code testing. testRigor generates tests from plain-English descriptions. NexGen QA differs by combining AI test generation and self-healing with security, database, performance, accessibility and LLM-validation modules, AI evaluation (tracing, red-teaming, human review) and a BYOK GitHub Actions runner in one workspace — see the detailed comparisons below for specifics.
Yes. Free is $0/month for 3 members with everyday QA, deterministic website checks and capped AI evaluation, and no credit card. Pro is $199 per workspace per month or $1,990 per year with 10 members included; extra members are $29/month or $290/year. All AI requires BYOK: your provider bills API usage separately, and there are no automatic overages. New prices apply to new subscriptions. Existing subscribers keep their current price, and we will give at least 30 days' notice before any change.
Still have questions? Talk to a solutions engineer.
Questions about Pro, a DPA, invoicing or on-premise deployment? Email us.
Email contact@qa-automation.com