Every serious test automation vendor now sells agents. They plan tests, write them, run them, sort the failures and fix what breaks, and Gartner gave the category its own Magic Quadrant in October 2026. This guide scores ten of them on what a team needs to know before buying: how much the agents do, whether they protect what a test is supposed to check, what they cover, and how hard they are to buy and to leave.
What counts as an AI testing tool in 2026?
The market splits three ways. Agentic platforms (mabl, Momentic, Functionize, Katalon, testRigor, Tricentis) run tests in their own format and add agents that create, heal and triage them. Managed services (QA Wolf) pair agents with human engineers and hand you the code. Layer tools sit on top of what you already run: Applitools adds visual checks to Playwright or Selenium suites, and Qodo works at the pull request and unit test level.
Gartner's first Magic Quadrant for Agentic Software Quality Assurance Platforms is dated October 6, 2026. Vendors announcing Leader placements include UiPath, TestMu AI (formerly LambdaTest) and Tricentis, which also says it placed furthest on completeness of vision; mabl announced a Challenger placement. The full vendor list is behind Gartner's paywall, so this guide does not reproduce it.
How do the tools score?
Four criteria, 1 to 5 each, from vendor documentation, announcements and pricing pages through October 31, 2026. These are editorial assessments of published capabilities, not benchmark results.
- Autonomy: how much of the plan, write, run, triage and repair loop the agents handle.
- Safety: how well the design protects test intent: deterministic execution, review of repairs, evidence on failures.
- Coverage: web, mobile, API, visual, performance and packaged enterprise apps.
- Access: published pricing, a free tier or trial, self-serve purchase, and whether you can take the tests with you.
| Tool | Autonomy | Safety | Coverage | Access | Total /20 | Model | Price (Oct 2026) |
|---|---|---|---|---|---|---|---|
| QA Wolf | 5 | 5 | 3 | 3 | 16 | Managed service + agents | Quote only (Coverage as a Service) |
| mabl | 4 | 4 | 4 | 3 | 15 | Low-code platform | Quote only; 14-day trial; 500 cloud credits/mo to start |
| Tricentis (Tosca, qTest, agentic platform) | 4 | 4 | 5 | 2 | 15 | Enterprise suite | Quote only |
| Functionize | 4 | 4 | 4 | 2 | 14 | Agent platform | Individual, Team, Enterprise plans; prices not published |
| Momentic | 5 | 3 | 3 | 3 | 14 | Agent platform | Free tier; paid plans quote only |
| Katalon | 3 | 3 | 4 | 4 | 14 | Platform + Studio IDE | Free plan; Professional per seat; Enterprise quote |
| testRigor | 3 | 3 | 4 | 4 | 14 | Plain-English platform | Free public plan; private from about $300/mo |
| Applitools | 3 | 4 | 3 | 3 | 13 | Visual AI + Autonomous | Free trial; paid plans quote only |
| Qodo | 3 | 4 | 2 | 4 | 13 | Code review + unit tests | 14-day trial; Pro Teams self-serve, credit-based |
| Testim (Tricentis) | 3 | 3 | 4 | 3 | 13 | Low-code SaaS | Free sign-up; paid plans quote only |
The tools
QA Wolf
QA Wolf sells end-to-end coverage as a service. Its agents plan, write, run, investigate and maintain tests in Playwright and Appium, start reviewing failures at the first red test rather than after the run, and repair and re-run tests before saving them. Human QA engineers work alongside the agents. Two design choices earn the top safety score: the output is deterministic Playwright code you own, and the company says it does not train models on customer data. QA Wolf publishes its own benchmark claims (93% on 1,567 evals against 78% for Claude Code) that have not been independently checked. Pricing is quote only. Best fit: teams that want broad end-to-end coverage without hiring a QA team. Full QA Wolf review.
mabl
mabl is a low-code testing platform for browser, mobile, API, performance and accessibility tests. Its April 2026 "Active Coverage" release added persistent agent instructions that apply team standards to authoring and recovery, cloud test generation from browser, CLI or IDE, runtime recovery, and a Jira and Confluence integration through Atlassian Rovo (PR Newswire). In October 2026 it added checks that run on a coding agent's change before a pull request is opened. mabl says its harness keeps business requirements intact when tests self-heal, which is the right problem to name. Prices are not published; a 14-day trial includes metered cloud credits, and local runs are free. Full mabl review.
Tricentis
Tricentis is the enterprise incumbent: Tosca for model-based automation, qTest for test management, plus NeoLoad, SeaLights and Testim. In March 2026 it launched an agentic quality engineering platform built around an AI Workspace that coordinates agents for test automation (now including SAP GUI), performance testing, test creation inside qTest and release-readiness analysis, with human review gates and audit trails in the execution path. It reported about $500 million in ARR for 2025 and more than 3,000 customers. Everything is sold through sales. Best fit: enterprises testing SAP, Salesforce and other packaged apps that need governance more than speed. Full Tricentis review.
Momentic
Momentic, founded in late 2023 and backed by Y Combinator and Standard Capital ($15 million Series A, November 2025), lets engineers describe flows in plain English and runs them as tests on web, Android and iOS. Its June 2026 release added an Explore agent that widens coverage and a failure classification agent that opens pull requests to fix false alarms (Newsfile). Beta figures it published: 600 tests written with a 73% merge rate. Customers named include Notion, Xero and Webflow. Paid prices are not published. Best fit: fast-moving product teams that want the most autonomous agents and will review their pull requests. Full Momentic review.
Katalon
Katalon pairs a desktop IDE, Katalon Studio, with a cloud platform. In April 2026 it launched True Platform, built on AWS and Amazon Bedrock, with agents for requirements analysis, test generation and maintenance, execution and bug detection, plus a natural-language assistant. It covers web, mobile, API and desktop and says more than 50,000 teams use it. Katalon publishes prices: a free plan, a Professional seat price and add-ons for runtime engines and device clouds; PricingSaaS recorded the Professional price moving during 2026, so confirm the current figure. Best fit: QA teams that want an IDE and a published price list. Full Katalon review.
testRigor
testRigor tests are written in plain English, which lets manual testers, analysts and product managers automate. It covers web, mobile, desktop and mainframe and now documents Claude Code and MCP integrations. Pricing is by parallel test infrastructure, not users or executions (testRigor); PricingSaaS recorded a free public plan and a private plan at $300 a month in June 2026. Best fit: teams whose testers do not write code. Full testRigor review.
Functionize
Functionize launched Functionize Studio in July 2026: agents generate tests against the live app, run them across browsers and update them when the app changes, while the tests themselves stay deterministic. It targets Salesforce, Workday, ServiceNow and SAP and raised a $41 million Series B in August 2025. Plans are Individual, Team and Enterprise, with prices not published. Best fit: enterprises with packaged apps that want an alternative to Tricentis. Full Functionize review.
Applitools
Applitools is two products. Eyes is visual AI: it compares screens against baselines and flags real visual differences. Autonomous writes tests from plain English, generates test data and runs on a deterministic engine rather than a live LLM. In September 2026 it shipped Eyes MCP tools for Claude Code, Cursor, GitHub Copilot and Cline, Figma design baselines, and plain-English steps for Playwright suites (TechIntelPro). Best fit: front-end teams that already run Playwright or Selenium and need visual regression checks. Full Applitools review.
Testim
Testim has been part of Tricentis since February 2022 and is sold as Tricentis Testim in web, mobile and Salesforce editions. It records flows into a low-code editor, uses AI locators that self-heal, and now includes Tricentis's agentic test automation and Testim Copilot for custom JavaScript steps. Sign-up is free; paid prices are not published. Best fit: product teams that want Tricentis support without Tosca's weight. Full Testim review.
Qodo
Qodo (formerly CodiumAI) raised a $70 million Series B in March 2026 and now describes itself as a code review and governance layer, with a multi-agent system that checks pull requests for bugs, compliance and architecture (SiliconANGLE). Its IDE plugin still generates unit tests. It is in this list because buyers search for it next to testing tools; it does not run end-to-end tests. Billing is credit-based with a 14-day trial and a self-serve Pro Teams plan (Qodo docs). Best fit: teams whose bottleneck is reviewing AI-written code before it reaches QA. Full Qodo review.
Also worth knowing
- Playwright test agents: planner, generator and healer agents in the open-source framework, run from VS Code, Claude Code, Codex or OpenCode. The healer skips a test it judges broken instead of forcing a pass.
- UiPath and TestMu AI: announced Leader placements in Gartner's agentic QA Magic Quadrant; consider them if you already run their automation or device cloud.
How should you choose?
- Decide who owns the tests. If you want code you can take elsewhere, QA Wolf or Playwright with Applitools. If you accept a vendor format for speed, mabl, Momentic, Katalon or testRigor.
- Match coverage to your apps. SAP, Salesforce and Workday point to Tricentis or Functionize. Web and mobile product teams have the widest choice.
- Ask about assertions. Can the agent change what a test checks during a repair? Require review for that, and ask for false-heal measurements on your own suite during the trial.
- Model the run cost. Credits, parallel slots and seats scale differently. Generated tests multiply run volume.
- Pilot on one low-risk suite. Measure escaped defects and weakened assertions, not tests generated.
Related questions
- What is agentic QA, and where does it fail? (partner site CTAIO)
- mabl, Testim or Tricentis: which fits an enterprise?
- QA Wolf, Momentic or testRigor: service or platform?
- Which AI code review tools are worth using?
- How do you evaluate an AI agent before production? (partner site CTAIO)
Sources
- UiPath, Gartner Magic Quadrant for Agentic Software Quality Assurance Platforms (report dated Oct 6, 2026); TestMu AI, Leader placement (Oct 13, 2026); BriefGlance, Tricentis Leader placement (Oct 9, 2026); mabl Challenger placement (Oct 2026).
- QA Wolf, Automation AI; PricingSaaS, QA Wolf pricing (captured Nov 2, 2026); AI Tools Atlas, QA Wolf review (verified Mar 2026).
- PR Newswire via WGN, mabl unveils next-generation agentic testing platform (Apr 23, 2026); mabl, Pricing; RuntimeWire, mabl adds pre-PR checks (Oct 21, 2026).
- Tricentis, Agentic quality engineering platform (Mar 10, 2026); 01net/Business Wire, Tricentis record year of ARR (May 7, 2026); Tricentis, Tricentis Testim; Testim, Testim and Tricentis announcement (Feb 2022).
- TechCrunch, Momentic raises $15M (Nov 24, 2025); Newsfile, Momentic launches agentic quality platform (Jun 23, 2026).
- Katalon, Katalon launches True Platform (Apr 2026); PricingSaaS, Katalon pricing (captured Nov 2, 2026).
- testRigor, How does testRigor's pricing model work; PricingSaaS, testRigor pricing (captured Jun 26, 2026).
- PR Newswire via Owler, Functionize launches Studio (Jul 16, 2026); Built In, Functionize raises $41M Series B (Aug 20, 2025).
- Applitools, Autonomous and Eyes: new AI features (Aug 7, 2025); TechIntelPro, Applitools visual AI guardrails for agentic coding (Sep 16, 2026).
- SiliconANGLE, Qodo raises $70M (Mar 30, 2026); Qodo, Pricing and usage.
- Playwright, Test agents; SD Times, The false-heal problem in AI test automation (Sep 2, 2026).
Related reading on our partner site: agentic QA rollout tiers for engineering leaders and AI code review policy on CTAIO.
What are the best AI testing tools in 2026?
On published capabilities, QA Wolf leads for teams that want a managed service, mabl and Tricentis lead for platforms you run yourself, and Momentic, Katalon, testRigor and Functionize follow closely. The right pick depends on whether you want to own the tests (QA Wolf exports Playwright code), test packaged enterprise apps (Tricentis, Functionize), or start cheaply (Katalon and testRigor publish entry plans).
What is agentic testing?
Agentic testing is test automation where AI agents run the loop: they explore an application, plan coverage, write tests, execute them, classify failures and repair broken tests, with people reviewing the output. Gartner published its first Magic Quadrant for Agentic Software Quality Assurance Platforms on October 6, 2026.
Is Testim still available?
Yes. Tricentis acquired Testim in February 2022 and sells it as Tricentis Testim, with editions for web, mobile and Salesforce. It now includes Tricentis's Agentic Test Automation, AI locators that self-heal, and Testim Copilot for generating custom JavaScript steps.
Is Qodo a testing tool?
Partly. Qodo, formerly CodiumAI, began with unit test generation and still generates tests from its IDE plugin, but since its $70 million Series B in March 2026 it positions itself as a code review and governance layer for AI-written code. It works at the code and pull request level, not end-to-end, so it complements the other nine tools here rather than replacing them.
Can AI testing tools replace Playwright or Selenium?
Some build on them rather than replace them. QA Wolf delivers Playwright and Appium code, Applitools adds visual checks and plain-English steps to Playwright suites, and Playwright itself now ships planner, generator and healer agents. Platforms such as mabl, Katalon and testRigor use their own test formats, which is faster to start and harder to leave.
Ready to Find the Right AI Tools?
Browse our data-driven rankings to find the best AI tools for your team.