Skip to main content
agent-qa on GitHub,230stars

Applitools compares pixels. agent-qa remembers behavior.

Applitools is strongest at visual AI validation, sold as an enterprise platform. agent-qa proves behavior: repo-owned natural-language flows, agent execution, hooks, and artifact review.

Try agent-qa, the source-available way to remember behavior, not just compare pixels.

Get StartedExplore Live Demo

agent-qa vs Applitools

Capabilityagent-qaApplitoolsDetails
Source accessApplitools is not positioned as a repo-owned framework with published source, so behaviour you disagree with is a support ticket. With agent-qa it is a pull request.
Repo-owned YAMLApplitools keeps the test intent inside its own product. agent-qa keeps intent, config, hooks, memory and suites beside the code they cover, where your engineering process already works.
Coding-agent nativeA coding agent cannot click through a hosted editor. agent-qa ships MCP tools, packaged Skills and a CLI, so the agent that changed the code writes the test, runs it and reads the failure without leaving the loop.
Bring your own LLMWhoever picks the model sets your quality ceiling and your bill. agent-qa lets you point at any provider, any compatible endpoint, or a model on your own hardware, and change it in one line.
Local and CI executionOne command on a laptop, in CI, and from an agent. No run depends on somebody else's control plane being up, and nothing queues behind another tenant.
Web and mobile QAWeb, Android and iOS from the same natural-language flow and the same evidence model. The surface is a target named in a file, not a different product tier.
Memory, cache, hooksExecution memory, a validated action cache and sandboxed hooks compound. A suite that has been running a month is faster, cheaper and better informed about your app than the day it was written.
No platform lock-inEvery durable asset stays in your repository. Cancel agent-qa tomorrow and the tests, the memory and the evidence are still there and still readable.

Why teams switch from Applitools

Runs compound instead of resetting

Applitools runs a test and forgets. Every run starts from nothing, which is why the hundredth run costs exactly what the first one did. agent-qa writes what it learned into memory committed beside your tests, so the next run starts where the last one finished and the suite gets better at your app on its own.

No license fee, no seat math

Applitools runs on enterprise platform pricing structured around visual checkpoints and seats, and that number grows with the coverage you add. agent-qa has no paid tier or licence fee for FSL-permitted use and its source is available under FSL-1.1-ALv2. You pay for tokens and infrastructure you control, on whichever provider is cheapest this quarter, and the cache cuts that too.

A pixel diff can't tell you the flow works

Visual validation catches what changed on screen; it can't tell you whether checkout completed, the API succeeded, or the right side effects fired. agent-qa verifies the behavior end to end and keeps screenshots as evidence, not as the definition of correctness.

Applitools answers 'does it look right?' agent-qa answers 'does it work?', and remembers the answer for the next run. Most teams need the second question answered first.

Frequently asked questions

Is agent-qa a good Applitools alternative?

Yes, and the reason is structural rather than a feature count. Applitools is a visual AI validation platform centered on screenshot comparison and visual regression, which means the asset you are building lives on their side of the line. agent-qa is a source-available QA agent with no paid tier, governed by FSL-1.1-ALv2: tests are plain-English YAML in your repository, runs execute on your laptop, in your CI, or from your coding agent, and every run writes back into memory committed beside the tests. The suite gets better at your app whether or not you renew anything.

How much does agent-qa cost compared to Applitools?

Applitools is priced on enterprise plans around visual checkpoints and seats, so the bill tracks how much you test. agent-qa has no paid tier, no seats and no platform fee; FSL-1.1-ALv2 governs permitted use. You pay for the model tokens and infrastructure you already control, on the provider you choose, and the validated action cache takes roughly 60% of the tokens off a matched rerun. Adding coverage does not add a line item.

How do I migrate from Applitools to agent-qa?

You are re-describing intent, not porting code, which is why this is far smaller than a normal test migration. Keep visual snapshots where they earn their keep; move the flow-correctness layer to agent-qa by writing your critical journeys as plain-English YAML tests. Run npx agent-qa init, write each critical flow as a plain-English YAML test, and let the runtime work out the selectors and the recovery. Most teams move a smoke suite in an afternoon, and there is nothing to un-pick later because the output is files in your own repository.

Does agent-qa cover web and mobile like Applitools?

Yes, and from the same file. agent-qa runs end-to-end tests on web, Android and iOS with one natural-language format, one memory store and one evidence model, so a flow written once survives being pointed at another surface. Applitools validates mobile visuals; agent-qa runs full mobile flows, taps, inputs, assertions, side effects, with visual artifacts attached to each step.

Can agent-qa do visual checks like Applitools?

agent-qa captures screenshots and artifacts at every step and can assert on visible state as part of a flow. It doesn't try to be a pixel-perfect visual regression engine. It makes behavioral correctness the contract and visuals the evidence, which is the priority order most teams actually need.

Sources

This page is based on public product and documentation sources. Verify current features and pricing with each vendor before making a purchase decision.

Where agent-qa pulls ahead of Applitools

The parts of agent-qa that answer what Applitools leaves you carrying.

Natural-language tests

Describe actions and assertions in natural language. agent-qa resolves them against the live interface using visible roles, labels, and screen state.

Learn about natural language tests

Natural-language YAML

Write the behavior and expected outcome in plain English. The test stays as reviewable YAML in your repository.

Targets users recognize

Refer to “New issue,” “Checkout,” or the “Issues table.” agent-qa finds the matching control in the live interface.

One format, every surface

Use the same natural-language structure across web, Android, and iOS without maintaining selector-heavy variants.

Self-healing execution

When a sub-action fails, because the click was intercepted, the control moved, or the field will not take the value, agent-qa re-observes the screen and plans another route to the same step, inside the same run. What it will not do is find another step: a recovery has to reach what you asked for, so a route to a different outcome is declined rather than taken, and the run fails with the reason it found.

Learn about self-healing

Re-observe, then re-plan

A failed sub-action is not retried. The screen is read again and the next attempt is planned against what is actually on it.

Bounded by the step

A recovery has to reach what the step asked for. A route that would reach a different outcome is declined, not taken.

The trace stays honest

Every attempt is recorded: what failed, what was tried, and what was refused. A healed step never reads as a clean one.

Version controlled, built for teams

Tests, suites, hooks, and the workspace config are files you write. What it learns about your app, the bugs it files, the rules you assert, and the skills an authoring agent uses are files agent-qa writes, in a visible directory beside your tests rather than in a database you cannot read. All of it is committed, so a new memory bundle arrives as a diff in a pull request with the evidence that justified it. Only derived indexes and binary run artifacts are gitignored, because they rebuild from what is committed. A teammate, a coding agent, and CI check out one commit and get the same brief.

Learn about configuration

Files, not a database

What you author and what the agent learns sit next to each other as committed files. Nothing it knows is locked in a store you cannot open.

Learning arrives as a diff

A new memory bundle or a filed issue shows up in a pull request, with the evidence behind it, and is approved the way any other change is.

Everyone reads the same commit

A teammate, a coding agent, and CI work from one checkout, so no run is quietly using state that nobody else has.

Bring your own model

The model is a setting, not a rewrite. Point a workspace at an OpenAI or Anthropic compatible endpoint, at Gemini, at an open-weight model running on your own hardware, or at a subscription your team already pays for like Codex or Claude Code, and override it on the one test that needs something stronger than the rest. Nothing about how a test is written changes when the model does, because the test says what should happen and the model is only what works out how to get there. There is no vendor to be locked to and no key of ours to buy.

Learn about LLM providers

Swap it in one line

The provider, the endpoint, and the model are config. Change them and every test you have already written runs against the new one, unedited.

The seat you already pay for

Account-backed authentication works where the provider allows it, so a Codex or Claude Code subscription can drive runs without a second bill.

Or nothing leaves the building

Point it at an open-weight model on your own hardware when the code under test is the kind that cannot go to somebody else's API.

* This comparison is based on publicly available information. Product capabilities and pricing can change; verify details with each vendor before making a purchase decision.