Agent Qa

What agent-qa does

agent-qa provides local MCP and CLI access for QA work across web, Android, and iOS projects. Its content tools focus on finding and reading the material needed to plan or support testing: official documentation, product pages, benchmarks, comparisons, articles, QA recipes, software factory guides, and company pages.

The server is suited to workflows where an AI agent needs grounded QA guidance rather than an isolated answer. Search results point to canonical or Markdown-readable pages, allowing the agent to inspect the underlying content before using it as evidence. The broader product description also includes persistent memidex memory and evidence-based self-improvement for ongoing QA work.

How it works

Run the local MCP and/or CLI workflow from a project configured for agent-qa. The MCP tools accept searches against different content collections and return ranked results or canonical Markdown URLs. A typical research flow is:

1. Search the official agent-qa documentation for product guidance.
2. Search Vostride content for relevant pages, benchmarks, comparisons, or QA material.
3. Search the QA library for a task-specific recipe or software factory guide.
4. Retrieve the selected page in Markdown to inspect the supporting content.

The server works alongside a browser or mobile test environment, depending on whether the target is web, Android, or iOS. Model access is also required for the agent workflow.

Setup and configuration

Use the free CLI with Node.js 24 or newer. A project configuration is required, along with access to a model and a suitable browser or mobile testing environment. The workflow can be used through a local MCP setup or the CLI.

Model and infrastructure expenses are separate from the CLI. Configure the project and its test environment before asking an agent to perform QA work so it has a target application and the required model access.

Tools and capabilities

• search_agent_qa_docs searches official agent-qa documentation and returns ranked matches.
• search_vostride_content searches titles, descriptions, and paths across product pages, benchmarks, comparisons, articles, documentation, QA recipes, and company pages. An omitted query can be used to browse.
• search_agent_qa_library finds task-specific QA recipes, software factory guides, and testing comparisons, returning canonical Markdown URLs.
• read_agent_qa_page retrieves the official Markdown form of a public Vostride content page.

The search tools are designed to be followed by read_agent_qa_page, so an agent can verify the page content behind a result.

Limitations and notes

The MCP tools center on searching and retrieving public documentation and QA-related content. A browser or mobile test environment remains a required part of the wider workflow. The package is offered under FSL-1.1-ALv2, which restricts competing commercial use. Model and infrastructure costs are not included.