A new full-stack Python framework named Plain just surfaced on Hacker News, explicitly engineered for interoperability with AI agents like Claude or Codex. Developers aim for code that is both explicit and predictable, ensuring smooth execution whether a human or an LLM interacts with the codebase. Getting started involves a simple `curl` command pulling down the starter script.

This framework forks from Django's architecture but reorients itself for agentic workflows, leveraging modern tooling like Python 3.13+, Postgres, htmx, and Tailwind CSS. Plain integrates agent tooling directly, featuring on-demand documentation accessible via CLI (`plain docs models`) and built-in "Skills" for complex tasks like package version upgrades or bug reporting, often triggering automated fixes.

Its structure emphasizes strong typing across its 30 first-party packages, covering everything from ORM (`plain.postgres`) to security scanning (`plain.scan`). The CLI commands, all run via `uv run`, offer developer conveniences like auto-reloading, linting across Python/CSS/JS, and integrated testing with pytest. Plain presents a cohesive, opinionated stack designed to minimize friction when agents modify or generate code.

This approach suggests a future where developer tooling proactively accommodates generative AI's integration into the software development lifecycle, treating agent interaction as a primary design constraint rather than an afterthought. PullApprove drives the ongoing development.