Our team kept running into the same thing: the coding agent would jump from a vague plan straight to code, and all the reasoning lived in a linear chat we couldn't revisit. So we built inplan.

It's a Markdown editor where you and the agent plan together in one doc: - Comment on a specific sentence and have a threaded conversation right there — ask, get the agent's opinion, decide — and it stays anchored to that line for later. - When the agent changes the plan, you review the diff and apply it, like a PR. - Installs as a Claude Code skill (npm i -g inplan); works with Codex and Pi too. Free and open source (AGPL-3.0).

Repo (with demo): https://github.com/melly-lgtm/inplan

Site: https://inplan.ai

Demo: https://github.com/melly-lgtm/inplan/releases/download/v0.1.18/inplan-demo.mp4

Built by our team — happy to answer anything. Feedback welcome, especially where it gets in the way.