Practical overview
When it fits
Use Agent Qa when you need natural-language end-to-end checks for web or mobile interfaces and want runs to re-observe the UI after a failed interaction. Its memory and action cache are relevant when later runs can benefit from prior observations or validated plans.
Who it is for
- Software developers writing or maintaining end-to-end tests for web and mobile products.
- Teams that want test actions and assertions expressed in natural language.
- Coding-agent users who want to invoke QA capabilities through the project's MCP integration.
Documented uses
- Define interface actions and assertions in natural language.
- Run web and mobile QA tests from a CLI or dashboard.
- Recover a test interaction by observing the changed interface and trying another path.
- Reuse observations and validated action plans across similar later test runs.
- Run setup, API, fixture, or teardown hooks in Docker containers during tests.
Requirements & limits
The repository identifies Agent Qa as open source and documents npm installation. Completing authentication is part of the dashboard setup flow; Codex or Claude Code subscription authentication uses a separately installed package.
- Browser setup documents Chromium support; mobile setup documents installing mobile drivers.
- Docker is required for the documented hook containers, not described as a prerequisite for every test run.
- The project lists OpenAI- and Anthropic-compatible endpoints, Gemini, local or open-source models, and Codex or Claude Code subscriptions as model/auth options.
- MCP support and coding-agent skills are included, but the repository does not detail setup for specific MCP clients.
- Tests, configuration, hooks, memory, and suite logic are described as version-controlled code.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Create and run a first web test
Define a YAML test for a configured web target and run that test from the CLI.
Before you start
- agent-qa initialized in the project; its default test directory is tests/**/*.yaml.
- A web target is defined in agent-qa.config.yaml with platform: web and a url.
- A stable test ID, which can be generated with the dashboard or CLI.
Steps
Add a named web target to agent-qa.config.yaml, including platform: web and its url.
Create a YAML file in the test directory with test-id, name, target, and steps; write the steps in plain language.
Run the test with npx agent-qa run followed by the test file path.
Expected outcome
The configured journey is run against the web target. The dashboard and CLI use the same file-backed test definitions.
Guide limitations
- The test file must refer to a configured target; the documented web target requires a URL.
- Do not put credentials in test context; the guide recommends environment files, secrets, or hooks for runtime values.
- This recipe covers a web test; browser and mobile configuration details are linked in the relevant documentation.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 16 October 2026
- Original website title
- vostride/agent-qa · Official homepage title · retained in its source language