Propolis – Browser Agents That Autonomously QA Your Web App

TL;DR

TL;DR: Propolis launches coordinated browser agents that explore your web app like real users, flag bugs, and auto-generate end-to-end tests you can drop straight into your CI pipeline. No manual test authoring required.

Source and Accuracy Notes

• Product: app.propolis.tech (YC X25)
• HN Launch: Launch HN: Propolis (YC X25) (116 points)
• Demo video: autonomous-qa-system-walkthrough
• Pricing: $1,000 per month for unlimited use with active support; lower-cost hobby plans available

What Is Propolis?

Propolis is a browser-agent-based QA platform that simulates real user behavior at scale. Instead of writing Playwright or Selenium tests by hand, you point Propolis at your web app and it fires a swarm of autonomous agents that collaboratively explore pages, identify friction points, and propose end-to-end tests derived from actual user journeys.

The founders — both with a decade of thinking about software quality — frame it as a “canary group” you can instrument without affecting real users. Deterministic tests are excellent for regression prevention, but they require you to know in advance what you’re testing. Propolis fills the gap where test coverage becomes guesswork.

How It Works

Step 1: Define Your Target

You provide the URL of the web app or specific flow you want tested. Propolis handles the rest — no test scripts, no configuration files.

Step 2: Launch a Swarm

Propolis spins up 10s to 100s of browser agents that work collaboratively. They navigate the app like real users, discovering:

• Broken interactions or dead-end flows
• Unexpected behavior in non-deterministic output
• Visual regressions and layout issues
• Friction points in multi-step workflows

Step 3: Receive Actionable Test Artifacts

The swarm returns a prioritized report of issues along with proposed Playwright or Selenium test snippets. These tests are not stubs — they are derived from real observed user journeys and can run directly in your CI pipeline.

Step 4: Integrate with CI

Generated tests are self-contained and portable. Drop them into your existing CI setup to run on every deploy, catching regressions that traditional unit tests miss because they mock too much.

Why This Approach Works

The core problem with hand-written E2E tests is that you only test what you expect to break. Stubbing and mocking remove the very real-world conditions that cause most production bugs. Propolis flips this: it observes what users actually do and generates tests for those paths.

Because the evaluation is LLM-assisted, Propolis can also catch quality issues in non-deterministic output — for example, a shopping assistant that recommends a product the user then searches for and cannot find.

Practical Evaluation Checklist

• [ ] Set up a free initial run (under 2 minutes)
• [ ] Observe how the swarm navigates your app’s key flows
• [ ] Review the generated test artifacts
• [ ] Run one or two of the proposed tests in CI against a known baseline
• [ ] Compare coverage against your existing Playwright test suite
• [ ] Evaluate pricing against your team’s QA engineering hours saved

Security Notes

• Propolis runs browser agents in sandboxed environments
• Test artifacts are generated from observed behavior, not from inspecting your codebase
• CI integration uses standard Playwright/Selenium test runners; no proprietary runtime required

FAQ

Q: Do I need to rewrite my existing Playwright tests to use Propolis?

A: No. Propolis generates new test artifacts alongside your existing suite. It is additive — you keep your deterministic regression tests and add coverage from agent-discovered journeys.

Q: How does it handle SPA navigation and dynamic content?

A: Browser agents operate in real browser instances with full JavaScript execution. They handle single-page applications and dynamic content just like a real user would.

Q: What happens if my app has authentication?

A: Propolis supports authenticated flows. You configure session state or credentials as part of the target definition, and agents navigate authenticated routes accordingly.

Q: Is $1,000 per month cost-effective for small teams?

A: For teams spending significant engineering time on test maintenance, Propolis can reduce that overhead substantially. There are lower-cost hobby plans for smaller projects — contact the team for caps.

Conclusion

Propolis addresses a real gap in the QA pipeline: coverage blind spots that hand-written tests cannot anticipate. By running swarms of browser agents that think like users, it surfaces issues before they reach production and generates runnable test artifacts you can keep.

If your E2E test suite is either too brittle to maintain or too sparse to catch real-world failures, Propolis is worth evaluating. The free initial run gives you enough signal to decide without committing to a paid plan.

Website: app.propolis.tech