Local threads

A Local thread runs an AI coding agent on your own Mac, in your project folder.

How a Local thread works

The desktop app starts the harness on your Mac, inside your project folder. The harness uses your own login and calls its model provider directly. Nothing goes through Cloudroom's servers.

flowchart LR
  subgraph mac [Your Mac]
    app[Cloudroom app] --> harness[Harness: Claude Code, Codex, or Pi]
    harness --> files[Your project folder]
  end
  harness --> model[(Your model provider)]

When to use Local

• You want the agent to edit the project folder you already have open in your editor.
• The task needs files, tools, or logins that only exist on your Mac.

Pick Cloud instead when the work should continue after you close your laptop, or when you want to run many agents without slowing your machine.

Work with a running agent

• Follow-ups: send a message while the agent works. On Local threads it steers the current turn.
• Stop: interrupts the current turn. Your history stays.
• Reasoning level: change it for the next message without starting a new thread.
• Many threads: run several agents at once, each in its own thread.

Move a thread to the cloud

Use Teleport to Cloud from the thread menu. Cloudroom stops the local agent, sends the conversation and project files to the cloud, and continues the same thread there.

sequenceDiagram
  participant Mac as Local thread
  participant Cloud as Cloud sandbox
  Mac->>Mac: Stop the local agent
  Mac->>Cloud: Send history and instructions
  Cloud->>Cloud: Resume the same conversation
  Mac->>Cloud: Upload project files
  Cloud-->>Mac: Done. Keep working from anywhere

Keep your laptop online until the completion notice appears. See Cloud threads for the way back.