# Optestra: full description

> Plain-English end-to-end tests for web, Android, CLI tools and Electron apps. AI writes each test once; every run after that replays it with no AI.

## What it is

Optestra tests websites, Android apps, CLI tools and Electron apps from plain-English test files (Markdown, stored in your repository). The first run records the steps with AI; every later run replays the recording with no AI. Each "Expect" line is compiled to a typed check that code evaluates, so a model never decides pass or fail. Failures come with the check's expected and found values, screenshots, video and a trace. When your app changes, a heal is proposed with its diff for you to review.

Open source (MIT): the engine, the test format, the CLI (`optestra`), the GitHub Action, the MCP server for coding agents and Bench. Closed: the cloud web app (hosted browsers and Android emulators, history, hosted inboxes, GitHub App, hosted AI, teams, billing). The desktop app comes after launch.

Status: The cloud web app is open.

## Pricing (from the pricing config)

1 credit = 1 cent. Local runs and writing tests with your own AI key: free, unlimited. A cloud run of 1–20 website tests: 2 credits (+1 per further 20). Android: 2 credits per test, at least 10 per run. A test written with our AI: 10 credits, +5 per 15 steps over 15 (a 40-step test: 20 credits); with your own AI key: 0. Heals: 0. "Explain this failure": 1. Cloud runs of CLI tools are coming soon (paid plans, priced like a website run); comparing two versions costs the two runs it contains, and is free on your own machine. Runs on your own machine, shown in your dashboard: free.

- Free: $0/month ($0/year), 100 credits (~10 tests written or ~50 cloud runs), 1 parallel runs, results kept 7 days
- Cloud (your own key): $5/month ($50/year), 1,000 credits (~500 cloud runs; bring your own AI key, so writing a test costs 0 credits), 2 parallel runs, results kept 14 days
- Solo Dev: $8/month ($80/year), 1,600 credits (~160 tests written or ~800 cloud runs), 2 parallel runs, results kept 14 days, +20% on top-ups
- Pro Dev: $19/month ($190/year), 4,000 credits (~400 tests written or ~2,000 cloud runs), 5 parallel runs, results kept 30 days, +25% on top-ups
- Team: $49/month ($490/year), 12,000 credits (~1,200 tests written or ~6,000 cloud runs), 10 parallel runs, results kept 90 days, shared workspace, +30% on top-ups

Top-ups: from $7, 100 credits per dollar, valid 12 months, with a bonus on paid plans. At zero credits: Free stops cloud actions; paid plans get alerts at 80% and 100% and a 10% grace buffer. Founding members: the first 100 subscribers keep their price for life and get 25% more credits. Machine-readable: https://optestra.com/pricing.json

## Questions

### Which AI sees my app, and what does it see?

Writing a test uses AI; replaying one does not. When Optestra's own AI writes your test, the request goes through OpenRouter to DeepSeek-V4.1-Flash, pinned to DeepInfra (US, fp8) with zero data retention, and with no fallback to any other provider. It receives the step text, a text description of the page and sometimes a screenshot. Secrets are typed by the test runner and never sent to the AI. If you use your own AI key instead, the request goes straight from the runner to your provider and not through us.

### Can the AI make a failing test pass?

No. Each Expect line becomes a typed check that plain code evaluates, and the verdict is computed from the checks. The AI works out how to do a step, never whether it passed. A heal can change how a step is done, never what is checked, and it is shown to you for review.

### What if you disappear?

Nothing breaks. Your tests are Markdown files in your own repository, the engine that runs them is open source (MIT) and runs on your machine or in your CI, and every recorded website test is also a plain Playwright spec that runs without Optestra. The cloud, the app and hosted AI are the paid parts; your tests don't depend on them.

### What does it cost?

Running tests on your own machine or CI is free and unlimited. In the cloud, 1 credit is 1 cent: a run of up to 20 website tests is 2 credits (2¢), and writing a test with our AI is 10 credits. The Free plan has 100 credits a month with no card, and paid plans start at $5 a month with your own AI key ($8 a month with ours too). Writing a test with our AI is 10 credits (5 more per further 15 steps over 15); with your own key it is 0. Every feature is on every plan, except cloud CLI runs, which are coming soon to paid plans, and there is no seat pricing.

### Playwright's test agents are free. Why would I pay?

You might not. Playwright's planner, generator and healer are good, free and open, and if your team is happy owning Playwright code they may be all you need. We are different in what you keep: the test is an English file anyone can read, each Expect is a checked-by-code assertion, heals are proposals you review, and the same file runs on Android. We also host the browsers and emulators, PR checks and inboxes, so there is nothing to set up. And because every website test exports as a Playwright spec, choosing us doesn't close the other door.

### Does every run use AI?

No. The first run works out the steps and records them. Later runs replay the recording with no AI, so a cloud run costs a couple of credits and the same result every time. AI comes back only for a new or changed step, or when your app changed and a step needs healing, and then it proposes a fix for you to review.

### How reliable is the AI at writing tests?

It depends on how you write. Step-by-step tests are the reliable path. On Bench's tidy, step-written shop tests the model authored 11 of 11 tests correctly, with 0 false passes and 1 wrong fail in 73 replays that should have passed. Messy prose and speech are worse: in the same run, a terse one-line description authored only 6 of 11. A wrong fail is annoying; a false pass is dangerous, so "zero false passes" is a claim we make only for step-written tests, not for prose or speech. Those runs are single runs, so read them as measurements, not guarantees.

### Do I need to write code?

No. Tests are plain English, one step per line, or you describe the test in a paragraph and let the app draft it. Engineers can add exact steps or Playwright code where they want to.

### Can it test localhost?

Cloud runs reach public sites and preview deploys, including protected previews. For localhost or a private network, run the open-source CLI or the GitHub Action on your own machine or CI runner, which is free. The Optestra desktop app, which also runs local tests, comes after launch.

### Is it safe to point an AI agent at my app?

The agent can only reach the domains you allow, enforced below the browser. It has a closed set of actions, no shell or file access, types secrets it never sees, and treats page content as untrusted data. Production environments block destructive actions unless a test declares them.

### Does it test Android apps?

Yes, in our cloud or on your own machine: your APK on a fresh emulator for every run, Android 13 to 17, with the same English test format. It costs 2 credits per test (at least 10 per run). iOS isn't supported.

### How do tests that need email or SMS codes work?

Email codes and links work with a hosted test inbox per project. SMS numbers aren't offered at launch. When a flow sends a code to your own phone, a run you are watching asks you for it in a six-box prompt and types it for you; when nobody is watching (CI, PR checks) the test is marked blocked, never failed and never charged.

### Does it work in CI and on pull requests?

Yes. Connect GitHub and every pull request is checked in our cloud with one comment that updates in place and a status check; a test that couldn't run is neutral, never a false pass or fail. Or run the open GitHub Action in your own runners with your own key. GitLab CI, CircleCI and Bitbucket use the same CLI.

### Can coding agents use it?

Yes: an MCP server, a CLI with JSON output, and instructions for agents that include the rule never to edit an Expect line to make a test pass.

### Where is my data kept, and can I delete it?

Results and evidence are kept for your plan's retention period (7 to 90 days) and then deleted. You can delete a project or your account and its data at any time, or ask us to. The sub-processors we use are listed with their purpose and location.

### Can it test CLI tools and desktop apps?

Web, Android, CLI tools and Electron apps, in our cloud or on your own machine. Native desktop apps are coming, on your own machine. You can also compare versions: run the same tests on two versions and get the regressions in one report. Native Windows, macOS and Linux apps are coming, on your own machine.

### Can I just buy credits instead of a subscription?

Yes. Stay on Free and top up from $7 when you need credits, at 100 credits per dollar, usable for 12 months. Compared to a plan you get fewer credits per dollar, one cloud run at a time, results kept 7 days, and no safety buffer when your credits run out.

## Bench (measured, with dates)

Replay with no AI, measured 2026-09-30 (engine f8a19d7, bench/baseline.json):
- Web shop (11 tests, 8 builds): false pass 0 of 15, wrong fail 0 of 73, flake 0 of 11 (10 reruns), 58 of 58 steps as recorded, 0 AI calls.
- Android app (7 tests, 6 builds): false pass 0 of 10, wrong fail 0 of 32, flake 0 of 7 (10 reruns), 34 of 34 steps as recorded, 0 AI calls.
After a purely cosmetic redesign of the web shop, 8 of 11 tests passed or healed with no AI and no re-recording, and 53 of 56 steps were done without AI. The rest need the fixer model. On the web shop, the replay's verdict and the generated plain-Playwright spec's verdict agreed on all 27 compared tests.

AI authoring (DeepSeek-V4.1-Flash through OpenRouter, pinned to DeepInfra fp8, zero data retention, no fallbacks), measured 2026-10-08 (engine 5ab6cf2). Each row is one run of the 11 shop tests, written in that style, authored by the model from scratch, then replayed with no AI on the 8 builds. "False pass" is a buggy build where the test passed (15 cases). "Wrong fail" is a correct build where it failed (73 cases).
- Tidy numbered steps (test file): authored 11/11, false pass 0/15, wrong fail 1/73.
- Mixed prose and steps (test file): authored 11/11, false pass 0/15, wrong fail 1/73.
- Given/When/Then (test file): authored 10/11, false pass 0/15, wrong fail 8/73.
- Terse developer notes (test file): authored 9/11, false pass 0/15, wrong fail 14/73.
- Verbose paragraph (test file): authored 9/11, false pass 0/15, wrong fail 14/73.
- Spoken transcript (test file): authored 9/11, false pass 0/15, wrong fail 14/73.
- Acceptance criteria (test file): authored 8/11, false pass 0/15, wrong fail 22/73.
- Sloppy (typos, wrong labels) (test file): authored 7/11, false pass 0/15, wrong fail 27/73.
- Verbose paragraph (description): authored 8/11, false pass 0/15, wrong fail 21/73.
- Terse developer notes (description): authored 6/11, false pass 0/15, wrong fail 34/73.
"Zero false passes" is claimed only for step-written tests. Wording matters: tidy steps author well; terse descriptions do not. Details: https://optestra.com/bench/

## Comparisons (every fact about others has a source and a date; see each page)

### Optestra vs TesterArmy (https://optestra.com/compare/testerarmy/)

A hosted AI testing agent with inboxes, SMS and iOS simulators, priced per run; against an open engine whose tests live in your repo and replay with no AI.
- Price. TesterArmy: Free 5 runs; Hobby $99/month for 250 runs; Startup $299/month for 1,000 runs; a hard stop at quota, no overage [source: https://docs.tester.army/account/plans-and-pricing.md, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- Cost per run, at a full quota. TesterArmy: About $0.40 on Hobby and $0.30 on Startup (our arithmetic: price ÷ runs) [source: https://docs.tester.army/account/plans-and-pricing.md, checked 2026-10-09]. Optestra: 2¢ for a run of up to 20 tests; replays use no AI.
- What counts as a run. TesterArmy: One test run, up to 20 minutes of execution, passed or failed [source: https://tester.army/pricing, checked 2026-10-09]. Optestra: One invocation of a set of tests in one environment; retries are free.
- Browsers and sizes. TesterArmy: Four named viewport presets: desktop, desktop-hd, mobile, tablet [source: https://docs.tester.army/run/viewport-size.md, checked 2026-10-09]. Optestra: Chromium, Firefox and WebKit; desktop, tablet and phone presets.
- Mobile. TesterArmy: iOS simulator and Android emulator in the cloud; no physical devices [source: https://docs.tester.army/mobile/overview.md, checked 2026-10-09]. Optestra: Android emulators (13 to 17) in the cloud, 2 credits a test; no iOS.
- Logins and codes. TesterArmy: Per-agent email inboxes and SMS numbers [source: https://tester.army, checked 2026-10-09]. Optestra: Saved logins, authenticator (TOTP) codes, email codes and links via a hosted inbox (or Mailpit, Mailosaur, MailSlurp), and a prompt for codes only a person can see; no SMS numbers at launch.
- CI. TesterArmy: GitHub Actions, GitLab CI and any pipeline that can call a webhook [source: https://tester.army, checked 2026-10-09]. Optestra: Open GitHub Action (one PR comment, a status check, sharding); GitLab, CircleCI, Bitbucket recipes.
- Export. TesterArmy: Docs list an export to TestRail as JUnit XML; we found no export of tests as code [source: https://docs.tester.army/llms.txt, checked 2026-10-09]. Optestra: Every website test is also a plain Playwright spec; standalone export.
- Open source. TesterArmy: The CLI is MIT; their GitHub organisation also lists an Apache-2.0 e2e framework (we haven't checked how it relates to the hosted product) [source: https://github.com/tester-army, checked 2026-10-09]. Optestra: Engine MIT; the app and cloud are closed.
Pick TesterArmy when: You need iOS simulators, SMS numbers or hosted inboxes today. You want a fully managed service now and don't need your tests as files in your repo. You'd rather pay per run with a hard cap than per credit.
Pick Optestra when: You want tests you own as files, that keep running as Playwright without us. You want repeat runs that cost cents and don't depend on a model. You want to run in your own CI for free, with your own key.

### Optestra vs Momentic (https://optestra.com/compare/momentic/)

A strong AI testing platform with a cloud step cache, visual diffs and mobile, priced in credits per step; ours replays from files in your repo and exports to Playwright.
- Where tests live. Momentic: Your repo, as YAML files [source: https://momentic.ai/docs/, checked 2026-10-09]. Optestra: Your repository: test files, recordings and generated specs are committed.
- Repeat runs. Momentic: A step cache, stored in their cloud per organisation, expiring 14 days after it was last saved [source: https://momentic.ai/docs/reliability/step-cache.md, checked 2026-10-09]. Optestra: Replays the recording with no AI; every check evaluated by code.
- Mobile. Momentic: Web on Chromium, iOS simulators, Android emulators; no physical devices [source: https://momentic.ai/docs/, checked 2026-10-09]. Optestra: Android emulators (13 to 17) in the cloud, 2 credits a test; no iOS.
- Visual regression. Momentic: Visual diff against golden screenshots [source: https://momentic.ai/docs/reference/commands/visual-diff.md, checked 2026-10-09]. Optestra: Not yet.
- Network mocking. Momentic: Yes: request mocking for web tests [source: https://momentic.ai/docs/platforms/web/request-mocking.md, checked 2026-10-09]. Optestra: Not yet.
- Price. Momentic: Free: 2,000 credits a month (their estimate: about 200 runs). Pay-as-you-go: $125 for 10,000 credits, then $0.01875 a credit (about 1,000 runs at their assumed 10 steps per run) [source: https://momentic.ai/pricing, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- What a credit is. Momentic: 1 per normal step, 2 per AI-generated or recovery step, plus browser (1 a minute) and Android (8 a minute) credits [source: https://momentic.ai/pricing, checked 2026-10-09]. Optestra: 1 credit = 1 cent; a cloud run of up to 20 tests is 2 credits.
- Seats. Momentic: No per-seat charges [source: https://momentic.ai/pricing, checked 2026-10-09]. Optestra: No seat pricing.
Pick Momentic when: You need visual regression, network mocking or iOS today. You want a hosted product with a large free tier now. Per-step credits with AI-assisted recovery suit how you work.
Pick Optestra when: You want the replay data in your repo, reviewable in a diff, with no expiry. You want to leave at any time with plain Playwright specs. You want a flat price per run instead of credits per step.

### Optestra vs QA Wolf (https://optestra.com/compare/qa-wolf/)

A managed service where agents and QA engineers write and maintain Playwright for you; we're a tool you run yourself, with English tests that compile to Playwright.
- Who writes the tests. QA Wolf: Their agents and QA engineers, as Playwright/Appium code [source: https://www.qawolf.com, checked 2026-09-26]. Optestra: You or your coding agent, in plain English; the first run records the steps.
- Repeat runs. QA Wolf: Deterministic code, no runtime LLM [source: https://www.qawolf.com, checked 2026-09-26]. Optestra: Replays the recording with no AI; every check evaluated by code.
- Export to code. QA Wolf: It is Playwright/Appium code [source: https://www.qawolf.com, checked 2026-09-26]. Optestra: Every website test is also a plain Playwright spec; standalone export.
- Mobile and desktop. QA Wolf: iOS, Android, Electron [source: https://www.qawolf.com, checked 2026-09-26]. Optestra: Android emulators (13 to 17) in the cloud, 2 credits a test; no iOS.
- Price. QA Wolf: Platform: 1¢ per AI credit and 15¢ per runner-minute. Managed Coverage as a Service: you pay for tests under management, on a custom quote [source: https://www.qawolf.com/pricing, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- Open source. QA Wolf: Their CLI is Apache-2.0; the platform is closed [source: https://github.com/qawolf, checked 2026-10-09]. Optestra: Engine MIT; the app and cloud are closed.
Pick QA Wolf when: You want people to write and maintain your tests for you. You need iOS or Electron coverage today. Budget matters less than handing QA off entirely.
Pick Optestra when: You want to write tests yourself, in English, and keep them. You want to run in your own CI for free. You want every fix shown to you for review, not made for you.

### Optestra vs testRigor (https://optestra.com/compare/testrigor/)

A hosted test platform for web, mobile and desktop apps, priced by parallel infrastructure with a public free plan; ours is priced per run, with private files in your repo.
- Platforms. testRigor: Ubuntu, Windows, Mac, Android, iOS and Windows native [source: https://testrigor.com/sign-up/, checked 2026-10-09]. Optestra: Android emulators (13 to 17) in the cloud, 2 credits a test; no iOS.
- Free plan. testRigor: Free plan: all tests and test results are public [source: https://testrigor.com/sign-up/, checked 2026-10-09]. Optestra: Free: 100 credits a month, no card, every feature except cloud CLI runs; your tests stay in your repo.
- Paid plans. testRigor: Private ("Complete") with a 14-day trial; Enterprise on request [source: https://testrigor.com/sign-up/, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- What you pay for. testRigor: Parallel test infrastructure: not users, not executions [source: https://testrigor.com/blog/faq-how-does-testrigors-pricing-model-work, checked 2026-10-09]. Optestra: Credits: 2 for a cloud run of up to 20 tests, 10 to write a test with our AI.
- Published price. testRigor: None: /pricing/ returned a 404 and the plan page shows no amounts [source: https://testrigor.com/pricing/, checked 2026-10-09]. Optestra: Yes: on this site and in pricing.json.
Pick testRigor when: You need iOS, Mac or Windows-native testing today. You want unlimited executions on a fixed amount of parallelism. You want a vendor-run platform and a sales conversation.
Pick Optestra when: You want to see the price before talking to anyone, and keep tests private by default. You want your tests as files you own, with a Playwright export. You want Android and web at cents per run.

### Octomind alternative (https://optestra.com/compare/octomind/)
Octomind has shut down. Per third-party coverage, it announced the end on 23 April 2026, switched the service off at the end of May and wound down at the end of June. We could not retrieve Octomind's own announcement: its website no longer resolves.
Octomind, the AI Playwright-test service, shut down in 2026. Its exports were plain Playwright code, so they still run; here is what to do next.
- Status. Octomind: Shut down: announced 23 April 2026, off at the end of May, wound down at the end of June (third-party coverage) [source: https://rywalker.com/research/octomind, checked 2026-10-09]. Optestra: The engine, CLI and cloud are all live.
- Last published plans. Octomind: Basic $89/month: 80 test cases, 240 cloud runs. Pro $589/month: 300 tests, 1,800 runs [source: https://rywalker.com/research/octomind, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- Your tests after the shutdown. Octomind: Exported as standard Playwright code [source: https://rywalker.com/research/octomind, checked 2026-10-09]. Optestra: Every website test is also a plain Playwright spec; standalone export.
- Open source. Octomind: Third-party coverage says its MCP server, CLI and GitHub Actions were MIT-licensed; the platform was not [source: https://rywalker.com/research/octomind, checked 2026-10-09]. Optestra: Engine MIT; the app and cloud are closed.
Pick Octomind when: Nothing to pick: the service has ended. If you still have its exported Playwright suite, keep running it with npx playwright test; it needs nothing from Octomind.
Pick Optestra when: You want plain-English tests that you own as files, after a vendor shutdown taught you why. You want to keep your Playwright specs and add English tests next to them (see the migration guide). You want cloud runs at cents, with the engine open source so a shutdown can't strand you.

### Optestra vs Playwright Test Agents (https://optestra.com/compare/playwright-agents/)

Playwright's free planner, generator and healer write and fix spec code in your editor; we keep an English test as the source, with typed checks, reviewed heals and verdicts.
- How tests are written. Playwright Test Agents: Markdown plans in specs/, generated .spec.ts files [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: Plain-English .test.md files, plus exact steps and Playwright code steps.
- Where tests live. Playwright Test Agents: Your repo [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: Your repository: test files, recordings and generated specs are committed.
- Repeat runs. Playwright Test Agents: Plain Playwright; LLM only when planning, generating or healing [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: Replays the recording with no AI; every check evaluated by code.
- Checks. Playwright Test Agents: Playwright expect and ARIA snapshots, as generated [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: Each Expect line compiled to a typed check, sanity-tested so it can fail; a model never decides the verdict.
- Healing. Playwright Test Agents: Healer replays the failure, inspects the UI, patches the code and reruns [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: No-AI fallbacks first, then a fixer model; every heal proposed with its diff and confidence, reviewed by default.
- AI models. Playwright Test Agents: Whatever your agent client uses (VS Code, Claude Code, Codex, opencode) [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: Our hosted AI (DeepSeek-V4.1-Flash, zero data retention), or your own key (Anthropic, OpenAI, Google, OpenRouter, Azure, Bedrock, any OpenAI-compatible or local model).
- Browsers. Playwright Test Agents: Playwright's browsers and device emulation [source: https://playwright.dev/docs/release-notes, checked 2026-10-09]. Optestra: Chromium, Firefox and WebKit; desktop, tablet and phone presets.
- Reports and CI. Playwright Test Agents: Playwright's reporters, traces and sharding; any CI [source: https://playwright.dev/docs/release-notes, checked 2026-10-09]. Optestra: Open GitHub Action (one PR comment, a status check, sharding); GitLab, CircleCI, Bitbucket recipes.
- For non-coders. Playwright Test Agents: No: code and an editor [source: https://playwright.dev/docs/test-agents, checked 2026-09-26]. Optestra: Desktop app: editor, live runs, results and fix review.
- Price. Playwright Test Agents: Free: Apache-2.0 and part of Playwright; you pay your own model provider [source: https://github.com/microsoft/playwright, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- Open source. Playwright Test Agents: Apache-2.0 [source: https://github.com/microsoft/playwright, checked 2026-10-09]. Optestra: Engine MIT; the app and cloud are closed.
Pick Playwright Test Agents when: Your team writes Playwright already and wants AI help inside the editor. You'd rather own and review spec code than English tests. You want nothing but Playwright in the stack.
Pick Optestra when: You want people who don't write code to write and read tests. You want checks that are proven able to fail, and heals you review before they land. You want verdicts, failure causes, a PR comment and costs per test out of the box.

### Optestra vs DIY Claude Code + Playwright MCP (https://optestra.com/compare/diy-claude-code-playwright-mcp/)

A coding agent driving a browser through Playwright MCP is great for one-off checks; tests need recordings, typed checks, verdicts and a way to run without the agent.
- What you get. DIY Claude Code + Playwright MCP: An agent session; code only if you ask for codegen [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: A test file, a recording and a verdict per run.
- Repeat runs. DIY Claude Code + Playwright MCP: A new agent session each time [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Replays the recording with no AI; every check evaluated by code.
- Pass or fail. DIY Claude Code + Playwright MCP: The agent's judgement, or browser_verify tools [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Each Expect line compiled to a typed check, sanity-tested so it can fail; a model never decides the verdict.
- Cost per run. DIY Claude Code + Playwright MCP: Tokens every run; accessibility snapshots can reach 100k+ tokens on a heavy page [source: https://provar.com/blog/thought-leadership/the-114k-token-problem-why-playwright-mcp-burns-your-ai-coding-agents-control-on-salesforce/, checked 2026-09-26]. Optestra: $0 AI on replays.
- Secrets. DIY Claude Code + Playwright MCP: --secrets dotenv values redacted from the model [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Typed by the harness into their allowed domains only; never shown to the model.
- Allowed domains. DIY Claude Code + Playwright MCP: Up to your setup [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Enforced below the browser: routes, a refusing proxy and a main-frame check.
- Evidence. DIY Claude Code + Playwright MCP: Session save, video, trace [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Screenshots per step, video with chapters, trace, network log, an HTML report.
- Coding-agent access. DIY Claude Code + Playwright MCP: It is the coding agent [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: MCP server, CLI with JSON output, AGENTS.md instructions.
- Price. DIY Claude Code + Playwright MCP: Free; you pay your model's tokens [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Engine, CLI and Action free; cloud from $5/month with your own AI key (1,000 credits; $8/month with our AI too, a run of up to 20 tests is 2¢), no seats.
- Open source. DIY Claude Code + Playwright MCP: Playwright MCP is Apache-2.0 [source: https://github.com/microsoft/playwright-mcp, checked 2026-10-09]. Optestra: Engine MIT; the app and cloud are closed.
Pick DIY Claude Code + Playwright MCP when: You want a quick look at a page while you code, not a lasting test. You're exploring, not checking the same thing on every change.
Pick Optestra when: You want the same check on every pull request, deterministic and free. You want a verdict you can trust without reading an agent's transcript. You want your coding agent to write real tests (through our MCP server) instead of improvising each time.

## AI E2E testing pricing (published prices, checked 2026-10-09)

- TesterArmy: Hobby $99/month for 250 runs; Startup $299/month for 1,000 runs. Free: 5 runs. Billed by: Runs (up to 20 minutes each); hard stop, no overage. Per run: $0.40 / $0.30 ($99 ÷ 250 runs; $299 ÷ 1,000 runs, at a full quota). Source: https://docs.tester.army/account/plans-and-pricing.md (checked 2026-10-09).
- Momentic: Pay-as-you-go $125/month for 10,000 credits; $0.01875 a credit after. Free: 2,000 credits a month. Billed by: Credits per step (1 normal, 2 AI or recovery) plus per-minute browser and emulator credits. Per run: about $0.125 ($125 ÷ 1,000 runs, their own estimate at 10 normal steps per run). Source: https://momentic.ai/pricing (checked 2026-10-09).
- QA Wolf (platform): Usage: 1¢ per AI credit and 15¢ per runner-minute. Free: Not stated on the page we read. Billed by: AI credits and runner-minutes; managed service by quote. Source: https://www.qawolf.com/pricing (checked 2026-10-09).
- testRigor: Not published. Free: Free plan: tests and results are public. Billed by: Parallel test infrastructure, not users or executions. Source: https://testrigor.com/sign-up/ (checked 2026-10-09).
- Checkly (Playwright monitoring, no AI): Starter $24/month for 3,000 browser runs; Team $64/month for 12,000 (billed annually). Free: Hobby: 1,000 browser runs, hard cap. Billed by: Check runs. Per run: $0.0065 (Starter overage, $6.50 per 1,000 browser runs). Source: https://www.checklyhq.com/pricing/ (checked 2026-10-09).
- mabl: Not published (request a quote). Free: 14-day trial. Billed by: Credits, from 500 a month; local runs free. Source: https://www.mabl.com/pricing (checked 2026-10-09).
- Reflect (SmartBear): Not published (contact sales). Free: 14-day trial. Billed by: Credits: web test 1, mobile test 5, API test 0.1. Source: https://smartbear.com/product/reflect/pricing/ (checked 2026-10-09).
- Autify Aximo / Nexus: Aximo Core $99/month billed annually ($120 monthly), 6,000 credits; Nexus Professional from $400/month. Free: Aximo: 2,000 one-time credits. Billed by: Aximo: credits per AI step (model multipliers); Nexus: per user, per parallel, per workspace. Source: https://autify.com/pricing (checked 2026-10-09).
- BrowserStack Automate: $59/month billed annually (1 parallel, desktop Chrome); $175 with mobile. Free: Not stated. Billed by: Parallel tests. Source: https://www.browserstack.com/pricing (checked 2026-10-09).
- TestMu AI KaneAI (ex-LambdaTest): Starter $17/month billed annually ($19 monthly), 2,000 credits. Free: None for KaneAI. Billed by: Credits per agent. Source: https://www.testmuai.com/pricing/ (checked 2026-10-09).
- TestSprite: Starter $19/month for 400 credits (first month free). Free: 150 credits a month. Billed by: Credits (the page doesn't define one). Source: https://www.testsprite.com/pricing (checked 2026-10-09).
- TestDriver: $20 per seat a month plus $0.14 per testing minute. Free: 14-day trial or 120 minutes, card required. Billed by: Seat plus testing minute. Source: https://testdriver.ai/pricing (checked 2026-10-09).
- Bug0: $2,500/month, managed, flat, up to 500 flows. Free: None; a discounted 60-day pilot. Billed by: Flat fee. Source: https://www.bug0.com/pricing (checked 2026-10-09).
- Maestro Cloud (mobile): $250 per device per month. Free: Local CLI and Studio are free and open source. Billed by: Devices (concurrent executions). Source: https://maestro.dev/pricing (checked 2026-10-09).
- Cypress Cloud: Team from $67/month; $6 per extra 1,000 test results. Free: 500 test results a month. Billed by: Test results. Per run: $0.006 ($6 ÷ 1,000 test results, overage). Source: https://www.cypress.io/pricing (checked 2026-10-09).
- Currents (Playwright dashboard): Scale $49/month for 10,000 results; $4.90 per extra 1,000. Free: None listed. Billed by: Test results. Per run: $0.0049 ($49 ÷ 10,000 results). Source: https://currents.dev/pricing (checked 2026-10-09).

## Security and data

Replays use no AI. Hosted AI: OpenRouter, pinned to DeepInfra (US, fp8) running DeepSeek-V4.1-Flash, zero data retention, no fallbacks. Your own AI key goes straight to your provider. Secrets are typed by the runner and never sent to an AI. Results and evidence are kept for the plan's retention (Free 7 days, Cloud (your own key) 14 days, Solo Dev 14 days, Pro Dev 30 days, Team 90 days). Customer data is processed in the United States. No cookies or analytics on this site. More: https://optestra.com/security/ and https://optestra.com/launch/#docs

## Subprocessors

- Google Cloud: Runs the app, the API and the browser and Android emulator workers; stores run evidence (screenshots, video, traces). (United States (us-east4, Virginia))
- Neon: Managed PostgreSQL database for accounts, projects, run records, the credit ledger. (United States)
- WorkOS: Sign-in (email, Google, GitHub) and sessions. (United States)
- Dodo Payments: Merchant of record: takes payment for subscriptions and top-ups, issues invoices, handles sales tax and refunds' payment side. (Dodo Payments' own locations (see their privacy policy))
- Resend: Sends our emails (usage alerts, monitor alerts, receipts) and receives mail for hosted test inboxes. (United States)
- OpenRouter: Routes our hosted AI requests (when you use our AI instead of your own key). (United States)
- DeepInfra: Runs DeepSeek-V4.1-Flash (fp8) for hosted AI. OpenRouter is told to use only this provider, with zero data retention and no fallbacks. (United States)
- Cloudflare: DNS and hosting of this website and the docs, and delivery of their files. (Global edge network)
- GitHub: Only when you connect a repository: reads your tests, posts the pull-request comment and the status check. (United States) [only if you connect it]

## Links

- Start free: https://app.optestra.com/sign-in
- Docs: https://optestra.com/launch/#docs (index for assistants: https://docs.optestra.com/llms.txt)
- Source: https://optestra.com/launch/#open-source