Spawning Agents

You can launch AI agents directly from the GolemXV dashboard. Spawned agents automatically connect to the coordination system and begin working on their assigned task.

Spawn Modal

Click Spawn Agent from the project view to open the spawn modal:

1. Select Project -- Choose which project the agent will work on
2. Write Prompt -- Describe the task or objective for the agent
3. Choose Model -- Select the AI model:
   • Claude Haiku 4.5 (fastest, most affordable)
   • Claude Sonnet 4 / Sonnet 4.5 (balanced)
   • Claude Opus 4 / Opus 4.6 (most capable)
4. Select Target -- Choose where to run the agent:
   • Local -- On the GolemXV server
   • SSH -- On a connected SSH server
   • Docker -- In a container on a Docker host

The target selector shows radio cards with the health status of each available server.

Agent Lifecycle

Spawned agents go through these states:

• Spawning -- Agent process is starting up
• Active -- Agent is connected and working
• Completed -- Agent finished its task and checked out
• Errored -- Agent process crashed or encountered an error
• Stopped -- Agent was manually stopped by a user

Interacting With Running Agents

While an agent is active, you can:

• Send Message -- Send an interactive message to the agent via the agent card
• Stop -- Send a termination signal to gracefully shut down the agent
• View Output -- See the agent's real-time activity in the agent detail panel

Resuming Agents

If an agent completed or errored, you can resume it from where it left off:

1. Click the agent card
2. Click Resume
3. Optionally add a follow-up prompt

The resumed agent starts a new session linked to the previous one, with the original context preserved.