Tejas Parthasarathi Sudarshan

Rocketman - Idea to production, managed in the repo.

A project management hub that lives in your codebase as one beautiful, offline HTML file, and a Claude skill stack that walks a product from raw idea to shipped, iterated reality.

Add a comment

Replies

Best
Tejas Parthasarathi Sudarshan
Hey Product Hunt 👋 I'm Tejas from Winsen Labs. We built Rocketman to fix the thing that kept biting us. The moment you start building with AI agents, your bottleneck stops being code. It becomes context. READMEs go stale. Tickets scatter across five tools. Every new agent session starts blind. It re-reads the repo, guesses the plan, reaches for the wrong library. Rocketman puts the whole project back where it belongs: in your repo. One offline HTML hub that lives in a PM/ folder inside every repo. Spec, kanban, roadmap, decisions, debug log, docs, and a live activity feed. All from structured JSON. All version-controlled. Double-click to open, see everything. No server, no SaaS, no login. And it's built for agents as much as for you: - A Claude skill track that walks an idea from ideate to PRD to plan to build to verify to test to launch. - An agent relay where multiple agents hand off work and talk to each other, every message and handoff visible in a Fleet view. - Decisions, bugs, and human vs agent provenance captured as you go. Nothing happens in a black box. We call it responsible token-maxing. Let Claude Code do the heavy lifting, but make every token leave a trace you can read, review, and trust. Here's the logic. On a long-running task, half of what an agent does is just maintain context. Re-reading the repo, re-establishing the plan, reloading state into every new conversation. By the time you're deep in, close to a third of the window is blocked holding state that never changed. Rocketman flips that. State lives in files. The agent loads only what the current task needs, when it needs it, instead of reloading everything every turn. No MCP calls, no tool-call overhead piling onto the context. The savings aren't a trick. They fall straight out of not paying for the same context over and over. Best proof we've got: the open-source Rocketman repo runs on Rocketman. The hub in the demo is the one we use to build it - https://winsenlabs.github.io/roc... Try it in one line: npx @winsendotai/rocketman init Free, MIT licensed, on npm today → github.com/winsenlabs/rocketman What would make this a no-brainer for your team? I'll be in the comments all day. 🚀