MCP/agent-qa
agent-qa logo

agent-qa

Verified Blue (500+ GitHub stars)

Self-improving AI QA testing harness with memory. Describe user journeys in plain language for web, Android, and iOS.

developer-tools•Testing & Automation
GitHub893
License:FSL-1.1-ALv2
VERIFIED
ByVostride@Vostride

About this Server

agent-qa is an open-source, self-improving QA agent and test harness with long-term memory for software engineering teams. Write end-to-end tests in plain English for web, iOS, and Android applications. Built with memidex, agent-qa learns from every test run, adapts dynamically to UI changes, records screenshots and DOM evidence, and prevents regressions before code ships to production.

Real-time Interaction

Direct low-latency RPC protocol communication with structured JSON schemas.

Scoped Sandboxing

Restricted capability bounds to prevent accidental modifications or leakage.

Security & Best Practices

Environment Isolation: Run agent-qa against dedicated staging or ephemeral preview environments rather than live production databases.
Credential Protection: Store test user credentials and API tokens in environment variables (.env.test) rather than plain test journeys.
Evidence Storage: Manage screenshot and trace storage policies to prevent sensitive user data capture in test artifacts.

When to Use (and When Not To)

Best For
  • •Writing and executing natural language QA tests for web, Android, and iOS user flows
  • •Validating checkout flows, authentication, and critical user funnels in CI/CD pipelines
  • •Self-adapting tests that maintain resilience against minor DOM or UI layout modifications
Not For / Limitations
  • •Direct database point-in-time recovery and snapshot rollbacks (use Rowsafe MCP instead)
  • •In-memory caching and Redis stream pub/sub management (use Redis MCP instead)
  • •Relational schema indexing and slow query optimization (use DBHub or PlanetScale MCP instead)

Install & Setup

terminal command
$claude mcp add agent-qa -- npx -y agent-qa mcp

Exposed Tools & Endpoints

5 tools

Functions your AI agent can invoke via this Model Context Protocol server

run_journey

Executes a natural language test journey on web or mobile platforms with visual and DOM step verification.

inspect_failure

Retrieves failure evidence, execution step traces, and screenshot comparisons for broken test runs.

query_memory

Queries memidex for remembered UI selectors, application knowledge, and working procedure procedures.

update_knowledge

Records validated procedure updates, selector revisions, and user feedback into repository knowledge files.

list_suites

Lists all defined test suites, user journey specifications, and historical pass/fail rates.