About Huddle
Real-time collaboration for the agent era.
We're building the layer that lets humans and AI agents work on the same code, in the same session, without stepping on each other.
Why we're building Huddle
A year ago, one of us was running Claude Code in a terminal while the other was firing off prompts in Cursor on the same repo. Within ten minutes the working tree was a mess of half-applied refactors, conflicting imports, and a frustrated merge commit nobody wanted to own.
We realised the entire industry had spent decades making Git great at async collaboration — but nobody had built the real-time layer for the moment when two humans and three agents all want to touch the same function at the same time. Google Docs solved this for writers in 2006. Developers deserve the same.
Huddle is that layer. A shared, live session where your teammate's edits, your agent's refactors, and your own keystrokes all land in the same document with cursors, presence, and conflict-free merges — across any IDE, any model, any host.
Principles
Your code lives on your machines. Huddle syncs state, not files. Go offline, keep working — the CRDT merges when you reconnect.
Every AI request uses your API keys, billed directly by the model provider. We never mark up tokens or resell compute.
Run Huddle on our cloud, on your own VPC, or on a laptop. The self-hosted tier is first-class, not a watered-down afterthought.
We build on MCP, CRDT, and open WebSocket specs. Lock-in is the opposite of our mission. If we vanish tomorrow, your tools still work.
Who's behind this
Huddle is a one-person operation out of Queensland, Australia. No VC, no team page full of stock photos — just a solo developer shipping a tool I wanted to exist. Every line of code, design call, and support email comes from me. If you want to say hi or report a bug — get in touch.