Skip to Content

Agent Management

Manage the full lifecycle of AI agents — create, configure, version, deploy, and monitor.

Agent List

The agents dashboard (/dashboard/agents) shows all agents with:

  • Name, model, mode (Reactive/Collaborative/Proactive)
  • Status badges — color-coded: green (active), yellow (updating), red (failed)
  • Quick actions: edit, duplicate, delete

Creating Agents

Three creation paths:

Basic Creator

/dashboard/agents/create — form-based creation with fields for name, model, system prompt, tools, and skills.

/dashboard/agents/templates — pre-built agent templates (Research, Code Review, Customer Support, etc.) that can be customized.

Agent Dev Environment

/dashboard/agents/create/agent-dev — advanced editor for building agents with full JSON config, testing, and iteration.

Agent Configuration

Each agent config includes:

FieldDescription
nameDisplay name
model_idLLM model (GPT-4o, Claude Sonnet, etc.)
system_promptBehavior instructions
toolsMCP tools the agent can use
skillsSkill categories
modeREACTIVE, COLLABORATIVE, or PROACTIVE
visibilityprivate, org, or public

Version History & Diff Viewer

/dashboard/agents/[id]/versions — view all config versions with timestamps and changed fields. Select two versions to see a side-by-side JSON diff with added (green), removed (red), and changed (yellow) highlighting.

Multi-Agent Specs

/dashboard/agents/multi — create multi-agent teams with DAG-based routing. Define entry agents, routing edges, and task dependencies.

Monitoring

Agent detail pages embed a monitoring section:

  • Grafana iframe when monitoring URL is configured
  • “Not configured” fallback with setup instructions