Launching today
spacecake

spacecake

Run Claude Code agents in terminal with a visual editor

33 followers

Visit website
Terminals are great for running agents, but terrible for editing markdown. spacecake is an open-source desktop app that supercharges Claude Code with a Notion-style markdown editor and integrated Ghostty terminal. The status bar shows your context window and usage cost ($) in real time, while the tasks panel gives you a live view of what agents are doing and what’s coming next.
spacecake gallery image
spacecake gallery image
spacecake gallery image
spacecake gallery image
Free
Launch tags:Developer Tools•Artificial Intelligence•GitHub
Launch Team / Built With
Threedium
Threedium
Image or Text to 3D Model
Promoted

What do you think? …

alexander
spacecake
Maker
📌
My workflow has completely changed in the last 6 months. I hardly ever write code by hand anymore - most of my time is spent writing markdown specs and prompts for Claude. The terminal is the best interface we have for agents, but it's a terrible interface for markdown. As a long-time Notion user suddenly living in the terminal, I wanted the best of both worlds. So I built spacecake.
Piroune Balachandran

Terminal handles the agent loop fine, but markdown streams look brutal. Been eyeing Claude Chic for the rendering side, though worktree support for parallel agents is what pulled me in there. Spacecake's visual editing angle could be the unlock if it handles concurrent sessions as cleanly.

alexander
spacecake
Maker

@piroune_balachandran thanks for the feedback!

You can track work in concurrent sessions through the tasks panel in spacecake (it shows which agent/subagent the task is assigned to).

It would be great to have a UI for worktrees though - I will add that ASAP.

Let me know if there are any other features you'd find useful.

Daniele Packard

Super cool! So you and claud code can collaborate on the md files and use them to inform future tasks?

alexander
spacecake
Maker

@daniele_packard absolutely!

My personal workflow is this:

  • Start Claude Code in plan mode (enabled by default in the spacecake terminal)

  • Iterate on a markdown plan (press ctrl+g to open it in the spacecake WYSIWYG editor)

  • If it's a small task/something to start right away, I press shift+enter for Claude to start building it

  • If it's a longer task or something to start later, I just save it to disk.

I also use markdown files for handoffs between agents - asking Claude to write a summary of references used, what's been built, what remains, and what the next priority is.

Then I clear the context window and have the next agent pick up the remaining work.

I find this works really well compared to auto-compaction, since you can easily review and edit exactly what's going to be loaded into the context window for the next agent.