Canary

Canary is an advanced quality assurance harness engineered specifically for AI coding agents such as Claude Code. By inspecting codebase changes and pull request diffs, Canary automatically identifies affected user interface flows, executes end-to-end browser tests via Playwright, and outputs complete session replays complete with video recordings, console logs, network HAR captures, and Playwright execution traces.

Key Features of Canary

• Claude Code Integration: Operates as a plugin, skill, or subagent workflow directly within the Claude Code terminal environment.
• Rich Session Capture: Records full screen video, browser console errors, network requests (HAR), and trace file archives during agent execution.
• Auto-Generated Playwright Scripts: Automatically converts the agent’s exploratory browser actions into clean, reusable Playwright scripts.
• Regression Replays: Replays captured test scripts in standard CI/CD pipelines with zero LLM inference costs.

Benefits of Using Canary

• Slashed Test Maintenance: Let the AI agent generate the E2E script once; replay it programmatically forever.
• Rich Visual Proof: Inspect failed automation runs with video timelines aligned to browser console and network request tracks.
• Context-Aware Scenarios: Focuses QA testing by identifying code modifications and targeting only the impacted UI components.

For engineering groups utilizing Claude Code, Canary bridges the gap between AI development and robust quality verification, ensuring that agent-driven changes are automatically translated into solid visual regression suites.