Changelog

July 12, 2026

• Connect Claude Code or any compatible MCP client with one URL: https://www.runnerqa.dev/api/mcp/mcp — no Agent API token required.
• Sign in with RunnerQA OAuth and approve explicit scopes on a redesigned consent screen; access is isolated per account and limited to the approved scopes.
• The agent manages projects, test cases, run evidence, failure analysis, and human-approved fixes — tests still execute locally through the CLI.

July 10, 2026

• Agent API responses now include next_steps with the exact CLI commands to run each test locally.
• Test case pages show a single “YAML is ready. Run this test locally.” card with copyable commands.
• The AI Integrations page walks through the full five-step agent flow, from creating a test to reading results.

July 9, 2026

• External AI agents can now list projects, create test cases, generate Maestro-compatible YAML, and read run results over a token-authenticated HTTP API.
• Agent tokens: created in the dashboard, shown once, stored only as hashes, revocable instantly.
• Rate limiting per token, with AI generation counted against the same per-user budget as the dashboard.

July 9, 2026

• Failure reports attach the screen's real accessibility ids and per-step timings, so AI fixes target real elements instead of guessing.
• AI fixes prefer accessibility ids and exact text over brittle selectors, and upgrade adjacent selectors opportunistically.
• Generated YAML uses settle timeouts and assert gates instead of blanket waits — passing tests got faster, not just greener.

July 8, 2026

• Dashboard: projects, plain-English test descriptions, AI-generated Maestro-compatible YAML with validation.
• RunnerQA CLI: pull tests, run them on your own simulator or device, report results back automatically.
• Run history with logs, durations, AI failure explanations, and one-click fixes.
• Dark mode, password recovery, and health checks.