Skip to main content
The Agents tab manages agent workspaces — isolated contexts with their own identity, tools, skills, and model configuration.

​
Agent Configuration

Each agent workspace includes:
  • Overview — workspace name, identity (name, emoji), default status
  • Model Selection — set primary model and comma-separated fallbacks
  • Files — list and edit agent workspace files
  • Tools — tool policy editor and profiles
  • Skills — per-agent skill allowlist
  • Channels — per-agent channel configuration
  • Cron Jobs — per-agent scheduled tasks

​
Multi-Agent Architectures

You can create multiple agents with different personalities, skill sets, and model configurations. Each agent can:
  • Handle specific channels (e.g., one agent for Telegram, another for Discord)
  • Specialize in domains (e.g., coding agent, research agent)
  • Use different models based on the task complexity
For advanced multi-agent setups, see OpenClaw docs: Multi-Agent.