Website - Docs - Discord - Bug reports
🧪 Generate and deploy end-to-end tests using just a prompt
Posium is an open-source AI testing platform that lets you generate end-to-end tests in minutes.
It's a full platform, from generation → execution → scheduling → debugging → maintenance.
- ✅ Browser testing: available now
- 📅 Mobile app testing: coming soon
- 🧩 BYOM (Bring Your Own Model): use your preferred LLM/VLM providers (or self-hosted models)
- End-to-end platform: generate, run, schedule, monitor, debug, and maintain tests
- AI-powered test generation (with human oversight): discover flows, plan scenarios, and write tests from real user journeys
- Flake-resistant by design: agents optimized for stability, accuracy, and repeatability
- Auth-ready workflows: supports passwordless flows like magic links and TOTP/OTP
- Run anywhere: CI/CD pipelines + scheduled runs
- Rich debugging artifacts: logs, screenshots, traces, videos, and step-level context
- Suite maintenance: update/repair tests as your UI evolves
- MCP + agent-native: Claude Code, Codex, Cursor, Gemini, etc. can create/update/run tests via MCP
If it runs in a browser, Posium can test it - any tech stack, fully end-to-end:
- SaaS products: onboarding, billing upgrades, role-based access, invite flows
- E-commerce: search → product details → cart → checkout → refunds/returns flows
- Marketing websites: forms, funnels, gated content, analytics-critical paths
- AI products & agents: chat workflows, tool-calls through a UI, long multi-step sessions, eval-style user journeys
- Internal tools: admin consoles, dashboards, ops workflows, CRUD-heavy apps
- Cross-app journeys: login → third-party auth → emails/OTPs → back to app
Posium generates tests by using your app like a real user in a browser:
- You describe scenarios (e.g., “sign up → upgrade plan → invite teammate → logout”).
- Posium launches a browser and navigates the app, observing the UI via:
- Screenshots + Vision models (VLMs)
- DOM + accessibility tree signals for reliable element targeting
- AI agents plan steps, execute them, validate outcomes, and convert the flow into a robust, repeatable test.
- Tests run in CI or on a schedule, and failures come with built-in debugging (traces, screenshots, video, logs) so you can quickly pinpoint what broke.
- 🧑💻 TypeScript
- ☘️ Next.js
- 🤖 Vercel AI SDK
- 🎨 TailwindCSS
- 🧑🏼🎨 shadcn/ui
- 🧪 Vitest
- 🎭 Playwright
- 🚀 Fastify
- 🔒 Better-Auth
- 🧘♂️ Zod
- 📚 Fumadocs
- 🌀 Turborepo
We'd love to connect with you!
- 💬 Join the Community: Chat with us on Discord
- 🐛 Report an Issue: Found a bug? Let us know!
- 📬 Contact Us: Have questions or want to partner? Reach out!
Released under AGPL-3.0.