Show HN: Hive – A full dev workspace (kanban/session modes+multi-repo+agent sdk)
1 points
1 hour ago
| 1 comment
| hive-ai.dev
| HN
I kept hitting the same problem with Claude:

The native Claude app is great but it can be much better when you unlock capabilities of desktop rather than the terminal. Such as:

- no task management

- no structure

- hard to work across multiple repos

- everything becomes messy fast

So I built a desktop app to fix that.

Instead of chat, it works more like a dev workspace:

• Kanban board → manage tasks and send them directly to agents

• Session view → the terminal equivalent of Claude code for quick iteration when needed/long ongoing conversations etc

• Multi-repo “connections” → agents can work across projects at the same time with context and edit capabilities on all of them in a transparent way

• Full git/worktree isolation → no fear of breaking stuff

The big difference:

You’re not “chatting with Claude” anymore — you’re actually managing work.

We’ve been using this internally and it completely changed how we use AI for dev.

Would love feedback / thoughts

It’s open source + free

GitHub: https://github.com/morapelker/hive

Website: https://morapelker.github.io/hive

▲
Gingiris1031
1 hour ago
[-]
The kanban-as-task-queue framing for AI agents is smart — it forces you to decompose work before handing it off, which tends to produce much better outputs than free-form chat. One challenge we ran into with multi-repo agent workflows is context window management when files across repos have conflicting conventions. Does Hive do anything to normalize context before feeding it to the agent?
reply
▲
moropex
1 hour ago
[-]
so the way hive handles cross repos, is by creating a temporary directory, with symlinks to the original worktrees. For all the model cares and knows, hes working on a mono repo of x projects, and is able to navigate it the way it knows best. All edits he does - are then propagated to the original worktrees in their original folders. So you kind of get the best of both worlds (without requiring you to feed the context manually)
reply