init

Initializing a multi-agent collaboration session involves creating an AgentHub session directory, generating a unique session ID, and storing the initial task configuration to coordinate multiple agents around a shared task.

An AgentHub session is a shared workspace for coordinating multiple agents around a single task. You need it when running refactoring, research, or evaluation workflows that require parallel agent approaches.

Yes, you can capture a baseline during setup by providing an optional eval command, metric name, and direction. This allows performance-focused sessions to automatically record baseline metrics for comparing parallel agent approaches.

Yes, AgentHub session initialization supports both interactive setup and argument-driven initialization. This allows you to launch task execution or research workflows using either guided prompts or direct configuration arguments.

You compare agent approaches by initializing a single AgentHub session with your task and agent count. The session captures the shared task context and optional baseline, enabling multiple agents to work from one starting point.

To start an evaluation-driven agent run, you specify the task, number of agents, optional eval command, metric direction, and base branch. The initialization process records these configuration values to prepare the parallel agent workflow.