Experiment: making multiple Claude AIs talk to each other

A user shared a setup where multiple Claude instances run at the same time and send messages to each other. The idea is that splitting roles between AIs — one writes, another reviews — can produce better results than using a single AI alone.

Normally, you run one Claude conversation at a time. This experiment connects several Claude instances so they can exchange messages with each other, like a small team. For example, one Claude writes code while another checks it for mistakes.

This approach is called a multi-agent setup. It lets complex tasks be split across several AIs that can catch each other's errors. For a solo developer handling a big project, this is a practical way to use AI like a team rather than a single assistant.

Key points

  • Multiple Claude instances run simultaneously and pass messages between each other
  • Roles can be split: one AI writes code, another reviews it
  • A multi-agent setup helps break complex tasks into parallel workstreams
  • Solo developers can use this to get team-like output from AI tools
  • The experiment is built inside the Claude Code environment

Quick term guide

share
A server folder made available to apps or other devices.
instances
Individual units of cloud computing power rented by the hour, such as a single GPU server slot.
instance
One independent run of an AI model, separate from any other run of the same model.
multi-agent
A setup where several AI agents each handle a different subtask and work together to complete a larger goal.
Solo developer
An individual who handles all parts of creating a project or product alone.
parallel
Doing multiple things simultaneously instead of one after another
developers
Developers are people who build software, apps, or websites.
AI tools
Software that can help create text, code, images, or other work.
Read original