Workflow · Production

Ace: GitHub Next's realtime multiplayer coding agent workspace for team alignment in agentic development

The problem

In early 2026, all coding agents are single-player experiences, but software development requires team alignment on what to build and why. As agents compress implementation time to minutes, existing coordination tools — GitHub, Slack, Jira, Linear — are not designed for the speed and volume of agentic work, making alignment the new bottleneck.

First attempt

Current tools surface alignment checkpoints at the PR stage — after implementation — rather than before it. Agent plan modes are unshared with teammates, and the result is merge conflicts from multiple agents touching the same files, duplicated work, and giant PR stacks with no context.

Workflow diagram · grounded in source
1
Session with shared cloud computer
trigger
“Each session is more than a chat channel. It is also backed by a microVM, so a sandboxed computer in the cloud on its own git branch.”
2
Multiplayer agent prompting
ai_action
“we can both prompt the agent. It's multiplayer prompting. The agent can also read our whole conversation - that's all input to the prompt.”
3
Live preview team review
human_review
“My team mates Nate and Idan are jumping in here They've taken screenshots, they've suggesting some alternative features, and asking questions.”
4
Agent writes feature plan
ai_action
“So here we're chatting about adding variable time frames to our calm hacker news app and I then ask Ace to make a plan.”
5
Team collaboratively edits plan
human_review
“I can see all my teammates cursors. We can collaboratively edit it together.”
6
Agent implements approved plan
ai_action
“once we're happy with all the details, we can go back to chat and ask Ace to build this and off it goes”
7
PR creation to GitHub
integration
“We can create this PR from directly inside Ace. We can open it on GitHub. It has a link back to our Ace session.”
8
Team dashboard and pulse
output
“We have access to lots of rich context on what work is underway and can helpfully summarise it”
Reported outcome

Ace is a research prototype entering technical preview that gives teams a shared multiplayer workspace — multiplayer chat, cloud-hosted microVMs, and shared coding agents — so planning and alignment happen continuously alongside implementation.

Reported metrics
Developer time reclaimed by agentsreclaiming your time
issue to agent PR timeonly a few minutes
Features shipped per developer per dayfive features a day instead of one
Reported stack
AcemicroVMOpus 4.6VS CodebunGitHub
Source
https://maggieappleton.com/zero-alignment
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Ace is a research prototype entering technical preview that gives teams a shared multiplayer workspace — multiplayer chat, cloud-hosted microVMs, and shared coding agents — so planning and alignment happen continuousl…

What tools did this team use?

Ace, microVM, Opus 4.6, VS Code, bun, GitHub.

What results were reported?

Developer time reclaimed by agents: reclaiming your time; issue to agent PR time: only a few minutes; Features shipped per developer per day: five features a day instead of one (source-reported, not independently verified).

What failed first in this deployment?

Current tools surface alignment checkpoints at the PR stage — after implementation — rather than before it.

How is this workflow AI workflow structured?

Session with shared cloud computer → Multiplayer agent prompting → Live preview team review → Agent writes feature plan → Team collaboratively edits plan → Agent implements approved plan → PR creation to GitHub → Team dashboard and pulse.