Data Flow

How information moves through Clustr when agents collaborate.

Agent spawn flow

  1. You click New Agent in the dashboard (or an existing agent spawns one programmatically)
  2. Clustr takes a git snapshot of your code for safe rollback
  3. Clustr launches the AI (Claude or Codex) in its own terminal with your task and collaboration tools
  4. Startup prompts and permissions are handled automatically
  5. The agent appears in the dashboard and begins working

Message flow

  1. Agent A sends a message to Agent B (by name or to everyone)
  2. Clustr saves the message and delivers it to Agent B's terminal
  3. The dashboard message feed updates in real time
  4. Agent B sees the notification, reads the full message, and can respond

Context flow

  1. An agent writes a note to shared context (e.g., key: "api-status", value: "migration complete")
  2. Clustr saves it and notifies the dashboard instantly
  3. Any other agent can read it at any time

Real-time updates

The dashboard stays in sync with the server through a live connection. Here's what updates automatically:

What changesWhen it updates
Agent listWhen an agent is spawned, starts running, finishes, or is stopped
MessagesWhen any agent sends a message
Shared contextWhen any agent writes or removes a context entry
Terminal outputAs agents produce output in their terminals
RulesWhen anyone edits the shared rules file
File changesWhen agents create, modify, or delete files in your project