TestivAI OSS
testiv.ai
- Free plan
- Self-hosted
What TestivAI OSS is
TestivAI OSS is a self-hosted tool for visual regression testing. It compares screenshots with DOM snapshots and examines pixel differences, DOM structure, computed styles, and element attribution to explain visual changes. Capture settings can freeze animations, hide the caret, and wait for web fonts. You can set a maximum difference percentage or pixel budget, and optionally pass rendering noise identified through DOM analysis. Integrations cover Playwright, Selenium, WebdriverIO, and Ruby Capybara/RSpec, with an experimental Java Playwright adapter. Cypress, Puppeteer, and Robot Framework use the Witness CLI on Chromium, also marked experimental. GitHub Actions can post reports to pull requests and support baseline approval through comments; documented workflows also cover GitLab CI, CircleCI, Jenkins, and Buildkite. An optional MCP server connects clients such as Claude Code, Cursor, or Copilot to structured results and explanations using your model. Reports are neutral checks by default, with optional hard failures. Data remains local. The MIT-licensed core is free forever, requires no account or API key, and costs nothing.
Who it suits
It suits developers and teams that need local visual regression checks in self-hosted or CI workflows. It is also suited to teams using the listed browser-testing frameworks or MCP clients.
Good to know, and what to watch
- Combines screenshot, DOM, style, and attribution analysis
- Supports Playwright, Selenium, WebdriverIO, and Capybara/RSpec
- GitHub Actions pull-request reports and baseline approvals
- Screenshots and diff images remain stored locally
- MIT-licensed core requires no account or API key
- Cypress, Puppeteer, and Robot Framework integration is experimental
- Java Playwright adapter is experimental
- Built by a solo developer without a company behind it
Plans
Every plan and priceTestivAI OSS
Free plan
The verdict
TestivAI OSS provides local visual comparisons with DOM-aware explanations and configurable tolerances. Its broad CI support and free MIT-licensed core fit teams that accept experimental integrations where noted.
What the maker publishes
- AI agents
- The optional MCP server connects Claude Code, Cursor, Copilot, or another MCP client to structured results and diff explanations using the user's model. source
- Analysis
- It combines pixel differences, DOM structure, computed styles, and element attribution to distinguish likely rendering noise from changes. source
- Capture
- Captures are stabilized by freezing animations, hiding the caret, and waiting for web fonts. source
- CI integration
- GitHub Actions can post visual reports to pull requests and support baseline approvals through a PR comment. source
- Frameworks
- Available integrations include Playwright, Selenium, WebdriverIO, and Ruby Capybara/RSpec, with an experimental Java Playwright adapter. source
- License and price
- The maker describes TestivAI as MIT-licensed, free forever, and usable in commercial products. source
- Maintenance
- The maker says TestivAI is built by solo developer Budi Sugianto and that there is no company behind it. source
- Notable limit
- Cypress, Puppeteer, and Robot Framework use the Witness CLI on Chromium; the maker labels the browser integration experimental. source
- Other CI
- The maker documents CI workflows for GitLab CI, CircleCI, Jenkins, and Buildkite using the same general capture and compare workflow. source
- Privacy
- Screenshots, DOM snapshots, and diff images are stored locally, and the maker says nothing is uploaded. source
- Purpose
- TestivAI performs visual regression testing by pairing screenshots with DOM snapshots and layered analysis to explain changes. source
- Report behavior
- Visual changes are reported as a neutral check by default, and teams can enable hard failures if they want them. source
- Support
- The maker lists [email protected] as its contact email. source
- Tolerances
- Users can set a maximum difference percentage or pixel budget and optionally auto-pass rendering noise identified by the DOM analysis. source
Checked against the maker's pages on 8 Oct 2026.