Hi HN — we built `dn` because coding got faster, but building software did not. Use the CLI to clear your backlog faster with reusable agentic workflows & other supporting commands. kickstart, meld, loop, land, fixup, and more commands can be used by you or your agent to turn markdown specification into implementation. Use the CLI in GitHub Actions overnight to durably handoff pull requests and github issues your team picks up in the morning.

For example:

```sh dn --agent codex kickstart --awp \ https://github.com/owner/repo/issues/123 ```

`dn` reads the issue and repository instructions, writes a named Markdown plan under `plans/`, asks the selected coding agent to implement it, and tracks the acceptance criteria. In AWP mode it creates a branch, commits, pushes, and opens the pull request.

You can install a dn skill for your agent. We currently support OpenCode, Claude Code, Cursor, Copilot, and Codex. The conversations you have with your agent are local to your machine. The `dn` skill helps record that context in Github issues and in your git repository as filesystem context. Productivity compounds when your team and the agents your team use share context.

The latest version adds Docker and exe.dev sandboxes for agent runs, remote progress reporting, and a clearer `meld` → `loop` → `land` workflow for splitting work across multiple sessions & developers. We'd love your feedback to make `dn` a better product, it's early days!

We also built https://denoise.cloud so product managers can build alongside `dn` users. Use denoise to short-circuit the software development lifecycle and unlock huge productivity gains for your team!. Our bet is that small teams do not need more dashboards or another proprietary agent. They need a clearer boundary between human judgment and machine execution.

See denoise in action at https://x.com/DenoiseCloud/status/2078607769259721016