Features
Auto-Fix Agent
FixSense's AI agent automatically creates pull requests to fix failing E2E tests.
Overview
The Auto-Fix Agent is an AI-powered workflow that automatically generates code fixes for failing Playwright and Cypress tests. It uses Claude to understand the failure context and create targeted fixes.
How It Works
- Trigger — When FixSense detects a failure with high fix confidence, it triggers the auto-fix workflow
- Analysis — Claude reads the failure analysis, test code, and application code
- Fix generation — The agent modifies the test or application code to resolve the failure
- PR creation — A new branch (
fixsense/fix-*) is created with a Pull Request
BYOK (Bring Your Own Key)
Auto-fix uses your own Anthropic API key. This means:
- Your code stays private — the agent runs in your GitHub Actions runner
- You pay Anthropic directly — standard API rates, no FixSense markup
- Full control — set your own spending limits in the Anthropic console
Monthly Limits
| Plan | Auto-fixes/month |
|---|---|
| Free | 3 |
| Pro | 20 |
| Team | 100 |
Auto-fix limits exist to prevent runaway usage. If you consistently hit your limit, consider upgrading your plan.
What the Agent Can Fix
- Broken selectors — updated locators to match new DOM structure
- Timing issues — added proper waits and assertions
- API changes — updated test data or mock responses
- Assertion updates — corrected expected values after intentional changes
What the Agent Won't Do
- Modify files outside the test or directly related application code
- Push directly to your main branch (always creates a separate branch)
- Run without your explicit API key configuration
- Access external services or databases
Security
The auto-fix agent operates within strict boundaries:
- Runs inside your GitHub Actions runner
- Only creates branches prefixed with
fixsense/fix-* - Never touches configuration files, secrets, or CI workflows
- All changes are reviewed via standard PR process