Until now, API work has been way more painful than it needs to be—especially if you’re building AI agents and you’re living in endpoints all day. The classic workflow is a mess: you build routes in your IDE (maybe with a coding agent), then you jump to another tool to create collections and tests, then you duct-tape it all into CI. Specs, collections, tests, and code end up scattered, not connected, and definitely not versioned together. In this video I walk through Postman’s massive update that basically fixes that. Postman is now git-native, so collections, specs, mocks, and tests can live directly in your codebase as simple YAML files. That means you can edit in Postman, flip to your editor, run a git diff, and review everything like normal code changes. The best part is Agent Mode: I can describe what I want in plain English, it scans my codebase (like a Flask app), discovers routes, generates specs, writes tests, and sets up mocks—then outputs reviewable git commits. You can choose models (Claude, GPT, self-hosted), pull context via MCP servers and tools like GitHub/Sentry/Linear, and run the exact same collection locally and in CI with pre-commit hooks and GitHub Actions for true local-to-CI parity.