# VirtualPyTest

> VirtualPyTest is an open-source, AI-native platform for testing and monitoring any device with a screen.

VirtualPyTest runs visual tests and 24/7 monitoring against real devices, including web browsers, phones, tablets, Android TV, smart TVs, set-top boxes, and desktop applications. It uses real device control, HDMI capture where applicable, image/OCR/AI-vision verification, execution evidence, and one reusable navigation model across platforms.

## Primary links

- Website: https://virtualpytest.com/
- Live read-only demo: https://virtualpytest.angelstreet.io/
- Source repository: https://github.com/AngelStreetCorp/virtualpytest
- Project README: https://github.com/AngelStreetCorp/virtualpytest#readme
- Documentation: https://github.com/AngelStreetCorp/virtualpytest/tree/main/docs
- AI agent guide: https://github.com/AngelStreetCorp/virtualpytest/blob/main/docs/ai/README.md
- MCP tool reference: https://github.com/AngelStreetCorp/virtualpytest/tree/main/docs/mcp
- REST API contracts: https://github.com/AngelStreetCorp/virtualpytest/tree/main/docs/api
- FAQ: https://github.com/AngelStreetCorp/virtualpytest/blob/main/docs/faq/README.md
- Overview PDF: https://virtualpytest.com/docs/VirtualPyTest-Overview.pdf
- Features PDF: https://virtualpytest.com/docs/VirtualPyTest-Features.pdf
- Contact: mailto:contact@virtualpytest.com

## Product areas

- Cross-platform visual testing for web, mobile, tablet, TV, set-top box, and desktop experiences.
- Navigation trees that separate reusable app navigation from test logic.
- Verification based on the actual rendered screen, with source frames, references, pixel differences, OCR, and AI vision detectors.
- Device fleet management, live screens, test execution, reports, monitoring incidents, and Grafana dashboards.
- REST API and CI/CD integration for triggering scripts and campaigns.
- AI-assisted test generation and an AI chat interface connected to project documentation.

## Agent execution

- MCP endpoint: `/server/mcp`, authenticated with the deployment's `MCP_SECRET_KEY` bearer token.
- Agent loop: initialize, list tools, take control of a selected device, inspect available actions, perform one action, capture or verify the resulting screen, then release control.
- CI/CD REST loop: `POST /server/script/execute`, poll `GET /server/script/status/{task_id}`, and retain the final result and evidence as CI artifacts.

## Licensing and commercial model

The software is open source under the GNU AGPLv3. Self-hosting is available at no software license cost. Hardware units and managed-lab services are separate commercial offerings; see the website for current details.

## Machine-readable companion

For a more detailed description, see https://virtualpytest.com/llms-full.txt.