Netlify Tests AI Coding Agents Before Launch Using Open Source AXIS Framework.

Netlify shipped a new way to run background functions, but the company didn't just push the update live and hope for the best. Before release, Netlify tested whether AI coding agents would choose the new pattern over the old one.

The deployment platform ran the same prompt across multiple AI coding tools, both with and without their custom skill loaded, to validate agent behavior ahead of time. The test spanned Codex, Claude Code, and Gemini, checking which execution path agents would follow when both options were technically valid.

Netlify's approach reflects a shift in how platforms think about feature releases. When both legacy and new patterns remain supported, the challenge becomes guiding AI agents to prefer the modern approach without forcing breaking changes on developers.

The company previously redesigned Netlify Functions for agent experience, introducing typed configuration and event handlers. With background functions, the same principle applies: maintain backward compatibility while steering agents toward better defaults.

Testing agent behavior before shipping ensures that the new pattern actually gets adopted. If agents consistently choose the legacy path despite updated documentation, the feature launch fails its primary goal.

Netlify announced that AXIS, the framework they use for agent testing, is now free and open source. The tool lets teams validate how AI coding assistants interact with their APIs and documentation before changes go live.

The testing workflow runs identical prompts in different environments. One test includes the updated skill or documentation; the control runs without it. Comparing outputs reveals whether agents pick up new patterns or fall back to familiar but outdated approaches.

For platforms targeting AI-native development, this kind of validation matters. Agents generate code based on training data, documentation, and context provided through skills or prompts. If the platform's guidance isn't clear enough, agents will keep using whatever pattern they learned first, even when better options exist.

Netlify's background functions update joins recent changes to function configuration and compute controls as part of the company's broader focus on agent experience. The platform has also added models like Claude Sonnet 5 to its AI Gateway and Agent Runners with zero-config access.

By open sourcing AXIS, Netlify gives other platforms a way to test agent behavior before launch. The framework addresses a specific problem: documentation alone doesn't guarantee agents will use new features correctly. Testing validates that guidance works in practice, not just in theory.

Details on AXIS and background functions are available through Netlify's official blog. The LinkedIn announcement didn't specify a repository URL or release timeline for AXIS, but confirmed its open source status.