Run a shared coding session with separate Claude Code agents

Let a group divide one feature across individual Claude Code instances, each with its own fresh context and review step.

Done with Claude Code· Advanced

What they did
The person made Claude Squad as a self-hosted place for group coding. To join, each participant runs one npx connect command that links their own Claude Code on their machine. An orchestrator agent then chops the feature into parallel pieces, sends each piece to a separate branch, and lines up the merges into one pull request; if a merge conflict appears, someone has to resolve it by hand.
How it went
The setup is hosted on free-tier Vercel, Supabase, and Partykit, with API keys staying on each participant’s computer and token costs spread across the group. The biggest limitation is still how well the work gets divided up, and the conflict flow plus after-merge review were still in an early state when the thread was posted.
Tips
Use this when you can write a clear spec first and split the work into independent chunks. Keep manual approval in the loop for any merge or other irreversible step, especially if conflicts are likely.

Try it yourself with Claude Code

Set up a collaborative coding session on [repo] for [number] people, where each person uses their own Claude Code from their own machine. Begin with a short spec for [feature], then break the work into separate tasks with fresh context for each one. Assign each task to its own branch, stop before any merge or submission, and ask me to approve anything irreversible such as merging, deleting, or sending. Once the branches are ready, run a review pass on each branch and list any conflicts or unresolved questions before proceeding.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on reddit.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.