# hive-mcp v0.11.0 — Stable Swarm with Consolidated Tools
The swarm finally works reliably end-to-end with fewer tools than last time.
## What's New
• *Consolidated tools*: 200+ MCP tools → 15 unified interfaces (`agent`, memory, kanban, magit, cider, hivemind, wave, olympus, agora, kondo, preset, project, session, emacs, kg)
• *Stable lifecycle*: spawn, dispatch, kill, and cleanup all work correctly
• *Project scoping*: memory and kanban no longer leak across projects (issue was happening while strangling towards tools-consolidation)
• *Olympus tracking*: lings show accurate working/idle status
## Key Fixes
• Kill flow now syncs DataScript with Emacs properly
• 42 duplicate tool registrations removed
• swarm_dispatch routes to correct handler with full preflight
• Kanban directory param for project scoping
• Event-type→slave-status mapping for correct Olympus display
## Breaking Changes
• *Kanban schema*: Tasks now use new schema format. Old kb- prefixed entries may need migration
• *Tool consolidation*: Old tool names deprecated (still work via compat shims, but prefer new consolidated names)
## Release
https://github.com/hive-agi/hive-mcp/releases/tag/v0.11.0
Announcements in this channel can be made once a month per project, see description
today I may launch the first working version of the Olympus web UI https://github.com/hive-agi/olympus-web-ui
Still not functional, but this is the WIP
Any thoughts on moving beyond Emacs- VS-Code - Neovim or just command line Claude Code?
@tbrooke Thanks for the interest! I would advise to see this https://youtube.com/live/Nl_rlj8KYPw?feature=share, as I show hive-mcp working. I think it becomes clear how Emacs is just a dependency. And that I actually use the claude-code terminal CLI to talk to the hivemind, not Emacs. And it's meant to be like that. Emacs is just a vessel the AI uses to control other instances of itself, etc. We don't use the editor as an editor, but as a platform for the AI. And this is something of a paradigm shift. It's not an AI pair programming tool or a Cursor kind of project. It's AI coordinating AI through the editor. Why Emacs? Because you have a lot of internals control without hitting wall I can inquiry about internal state. Are claude instances stuck on a prompt? Are they idle? Are they working? As well as the TCP runs there such that it makes possible the piggyback system. There is an install video from a fresh ubuntu, if you get interested from seeing the demo: https://youtu.be/eE8t9IWD94I That being said, I don't plan to make that layer run anywhere else. But, I will port Olympus which is platform agnostic to multiple UIs: web, maybe https://github.com/charmbracelet/crush which is very popular, some GUI (probably done with a clojure library), etc. In Olympus you can basically see in a "dashboard view" the swarm working. The view logic runs in Clojure. The UI is a basic wrap binding that could be anything that support showing terminals running. I may address this in a separate video, since it may get a lot of people confused. And repel them because they think "oh, this is an Emacs project". It isn't.
Cooked up another take on map munging, thanks for checking it out https://github.com/tailrecursion/restructure
> but keeps the shape in one place What do you use to change shape?
By shape there I mean the selector argument
oh, I thought it's about the result. But it seems like this library preserves shape of input.
Not entirely since you can remove things. Depends on meaning of shape I suppose
ah, understood. There is a library which seems similar: https://github.com/turtlegrammar/faconne
Not familiar with that one but there's a lot of similar things, I'm aware of also Specter and Traversy
Meander also